Hasselmann Etal DHZ1973

Measurements of Wind-Wave Growth and Swell Decay during the Joint North Sea Wave Project (JONSWAP)
U D C 551.466.31; A N E German B i g h t
By K . Hasselmann, T. P. Barnett, E.Bouws, H.Carlson, D . E. Cartwright, K . Enke, J. A. Ewing, H . Gienapp, D . E. Hasselmann, P, Kruseman, A. Meerburg, P. Miiller, D . J. Olbers, K . Richter, W. Sell, H . Walden
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift Reihe A (8), N r . 12
1973 DEUTSCHES HYDROGRAPHISCHES INSTITUT HAMBURG
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift

Herausgegeben vom Deutschen Hydrographischen Institut Reihe A (8), N r . 12
U D C 551.466.31; A N E German B i g h t
1973
DEUTSCHES HYDROGRAPHISCHES INSTITUT HAMBURG
U D O 551.466.31; A N E German B i g h t
By K . Hasselmann, T. P. Barnett, E. Bouws, H . Carlson, D. E. Cartwi-ight, K . Enke, J. A. Ewing, H . Gienapp, D.E.Hasselmann, P. Kruseman, A.Meerburg, P.Mller, D. J. Olbers, K . Richter, W. Sell, H . Walden
1973
DEUTSCHES HYDROGRAPHISCHESINSTITUT HAMBURG
Ansohriften der Vertasser
K. Hasselmann, Institut fr Geophysik, Hamburg, Germany; 1970-72 also Doherty Professor at Woods Hole Oceanographio Institution, Woods Hole, Mass., U.S.A. T. P. Barnett, Westinghouse Ocean Research Laboratory, San Diego, Calif., U S A - now at Scripps Institution of Oceanography, La Jolla, Calif., U.S.A. B. Bouws, Koninklijk Nederlands Meteorologisch Instituut, De Bilt, The Netherlands H . Carlson, Deutsches Hydrographisches Institut, Hamburg, Germany D . E . Cart-iwight, National Institute of Oceanography, Wormley, Great Britain K. Enke, Institut fr Geophysik, Hamburg, Germany J. A. Ewing, National Institute of Oceanography, Wormley, Great Britain H . Gienapp, Deutsches Hycbographisches Inatitut, Hamburg, Germany D. E. Hasselmann, Institut fr Geophysik, Hamburg, Germany P. Kruseman, Koninklijk Nederlands Meteorologisch Instituut, De Bilt, The Netherlands A. Meerburg Koninldijk Nederlands Meteorologisch Instituut, De Bilt, The Netherlands; now at Dutch Ministry of Foreign Affairs, The Netherlands P. Mller, Institut fr Geophysik, Hamburg, Germany D. J. Olbers, Institut fr Geophysik, Hamburg, Germany K . Richter, Deutsches Hydrographisches Institut, Hamburg, Germany W. Sell, Institut fi' Geophysik, Hamburg, Germany H . Walden, Deutsches Hydrogi'aphisches Institut, Hamburg, Germany
Deutsches Hydrographisches Institut, Hamburg 1973 Schriftleitung: Dr. Helmut Madler, 205 Hamburg 80, Sichter 4 und Walter Horn, 2 Hamburg 6, Pehx-Dahn-StraBe 2 Druck: J. J. Augustin, Glckstadt/Blbe
CONTENTS Summary - Zusammenfassung - Rsum Part 1. The experhnent 1.1. Introduction 1.2. The wave profile 1.3. Other field measui-ements 1.4. Intercomparison of frequency wave spectra 1.5. Dh'eotional measurements 1.6. Cases studied 1.7. Logistics Aclinowledgments Part 2. Wave growth 2.1. 2.2. 2.3. 2.4. 2.5. 2.6. 2.7. 2.8. 2.9. 2.10. 2.11. Generation cases Kitaigorodskii's similarity law W i n d stress and turbulence measiu-ements Fetch dependence of one-dimensional spectra The dhectional distribution The mean som-oe function Analysis of individual generation cases Origin of tlie spectral peali The energy and momentum balance of the spectrum Variability of the spectrum Conclusions 7 10 10 15 16 19 19 22 25 26 27 27 28 28 32 38 39 42 48 49 52 57 71 71 73 74 81 87 89 90 02 94
Part 3. Swell Attenuation 3.1. 3.2. 3.3. 3.4. 3.5. 3.6. Conservation of action and energy flux Attenuation due to bottom friction Energy flux analysis of swell data Lateral variations of bottom topography Dispersion characteristics Conclusions Appendix: Statistical analysis of the attenuation parameter F
Notation References
Summary: Wave spectra were measm-ed along a profile extending 160 kilometers into the North Sea westward from Sylt for a period of ten weeks in 1968 and 1969. During the main experiment in July 1969, thirteen wave stations were in operation, of which six stations continued measiu'ements into the first two weeks of August. A smaller pilot experiment was carried out in September 1968. CWrents, tides, air-sea temperatme differences and turbulence in the atmospheric bomidary layer were also measured. The goal of the experiment (described in Part 1) was to determine the structm-e of the source fimction governing the energy balance of the wave spectrum, with particular emphasis on wave gi-owth imder stationary offshore wind conditions (Part 2) and the attenuation of swell in water of finite depth (Part 3). The som'ce fmictions of wave spectra generated by offshore winds exhibit a characteristic plus-minus signature associated with the shift of the sharp spectral peak towards lower frequencies. The two-lobed distribution of the source function can be explained quantitativeljr bj' the nonlinear transfer due to resonant wave-wave interactions (second order Bragg scattering).'The evolution of a pronomiced peak and its shift towards lower frequencies can also be understood as a selfstabilizing feature of this process. Por small fetches, the principal energy balance is between the input by wind in the central region of the spectrum and the nonlinear transfer of eiiergj' away from this region to short waves, where i t is dissipated, and to longer waves. Most of the wave growth on the forward face of the spectrum can be attributed to the nonlinear transfer to longer waves. For short fetches, approximately (80+ 20) % of the momentum transferred across the airsea interface enters the wave field, in agreement with Dob son's direct measurements of the work done on the waves by surface pressirres. About 80-90% of the wave-induced momentum flux passes into currents via the nonlinear transfer to short waves and subsequent dissipation; the rest remains in the wave field and is advected away. A t larger fetches the interpretation of the energy balance becomes more ambiguous on account of the milmown dissipation in the low-frequency part of the spectrum. Zero dissipation in this frequency range yields a minimal atmospheric momentum flux into the wave field of the order of 20 % of the total momentmn transfer across the air-sea interface - but ratios up to 100 % are conceivable i f dissipation is important. I n general, the ratios (as inferred from the noiflinear energy transfer) lie within these limits over a wide (five-decade) range of fetches encompassing both wave-tank and the present field data, suggesting that the scales of the spectrum continually adjust such that the wave-wave interactions just balance the energy input from the wind. This may explain, among other featiu'es, the observed decrease of P h i l l i p s ' "constant" with fetch. The decay rates determined for incoming swell varied considerably, but energy attenuation factors of two along the length of the profile were tjrpical. This is in order of magnitude agi'eement with expected damping rates due to bottom friction. However, the strong tidal modulation predicted by theory for the case of a quadratic bottom friction law was not observed. Adverse winds did not affect the decay rate. Computations also rule out wave-wave interactions or dissipation due to turbulence outside the bottom boundary layer as effective mechanisms of swell attenuation. We conclude that either the generally accepted friction law needs to be significantly modified or that sorne other mechanism, such as scattering by bottom u-regularities, is the cause of the attenuation. The dispersion characteristics of the swells indicated rather nearby origins, for which tho classical 5-event model was generally inapplicable. A strong Doppler modulation by tidal cm'rents was also observed.
Messuiigen des Wind-Wellen-Waehstums uiid des Dnungszeralls whreiid des Joint North Sea
Wave Project (JONSWAP) (Zusammenlassung). Walxrend einer Zeitspanne von insgesamt zelm Woohen in den Jaliren 1968 und 1969 wurden langs eines Profils, das sich 160 km weit seewarts westlich Sylt in die Nordsee hinaus erstreckte, Seegangsspektren gemessen. Walirend des Hauptexperiments im Juli 1969 arbeiteten dreizelm WellemneBstationen, an sechs Stationen wurden die L'lessmigen audi in den ersten zwei Augustwoohen weitergefiUu't. Ein kleineres Vorexperiment wtu'de im September 1968 ausgeflirt. Strmungen, Gezeiten, Wasser-Luft-Temperaturdifferenzen und Tm'bulenz in der atmospharischen Grenzschicht wurden zusatzlich gemessen. Das Ziel des Experiments (besclirieben in Teil 1) war es, die Quellfunktion zu ermitteln, die die Energiebilanz des Seegangsspektrums bestimmt, wobei besonders das Wellenwaohstum bei stationaren ablandigen Winden (Teil 2) mid die Dampfung der Dnung bei encUicher Wassertiefe imtersucht -ivm-den (TeU 3). Die Quellfimlition des von ablandigen Winden erzeugten Seegangsspektrums zeigt erne charakteristische Plus-Minus-Signatiu, verbunden mit einer Verschiebung des scharfen spektralen Peaks zu niedrigeren Frequenzen. Diese zweiteilige Strulitur der Quellfunlrtion kaim quantitativ
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12, 1973
durch nichthneare Energieiibergange infolge resonanter Wellen-Wellen-Wechselwirkungen (BraggStreuung zweiter Ordnung) erklart werden. Die Entwicldung eines betonten Peaks uird seine Verschiebung zu niedrigeren Frequenzen kann audi als Selbststabilisierung dieses Prozesses verstanden werden. Fr kurzen Fetch wird die Energiebilanz im wesentlichen ausgeglichen zwisclien der Zufulu- durch den Wind im Hauptteil des Spektrmns und dem nichtlinearen Transport von Energie aus diesem Bereich zu krzeren Wellen, wo sie dissipiert whd, und zu langoren Wellen. Der Hauptanteil des Wellenwaclistums an der Vorderflanlie des Spektrums kami auf nichthneare tibergange zu langoren Wellen zurckgeflirt werden. Fr km"zen Fetch flieBen (80 + 20) % des Impulses, der an der Grenzflache Luft-Wasser bertra-gen wird, in das Wellenfeld; das steht in bereinstimmung mit Dobson's d'ekten Messungen der Arbeit, die vom Oberflachendruok an den Wellen geleistet wird. Ungefalu' 80-90% des welleninduzierten Impvilsflusses wird ber nichtlinearen Transport zu kurzen Wellen und folgende Dissipation an die Strmung abgegeben, der Rest bleibt im Wellenfeld mid wird von diesem fortgefhrt. Bei langerem Fetch wh-d die Interpretation der Energiebilanz wegen der unbekannten Dissipation im niederfrequenten Teil des Spektrums ungewisser. Keine Dissipation in diesem Frequenzbereicli liefert eine imtere Grenze / 4- 20 des atmospharischen Impulsflusses in das Wellenfeld von etwa 20 _ % des gesamten Impulsflusses durch die Grenzflache Luft-Wasser - aber Werte bis 100% sind vorstellbar, wenn die Dissipation wesentlich ist. I m allgemeinen liegen die Werte (bestnmt aus den nichtlinearen Energiebergangen) innerlialb dieser Grenzen, und zwar fi' einen weiten Bereich (fnf Dekaden) des Fetch, wobei sowohl Daten aus Wellentanks als auch von den hier dargestellten Feldmessungen eingeschlossen sind. Dadurch wu-d nahegelegt, daB sich die Skalen des Spektrums kontinuierhch derart einstellen, daB die Wellen-Wellen-Wechselwirkungen gerade die Energiezufulir dm'ch den Wind ausgleichen. Dies kann imter anderem die beobaclitete Abnalnne der Phillipssohen ,,Konstante" erklaren. Die Zerfallsraten, die fr einlaufende Dnung bestimmt wurden, variierten betraehtlich, aber Energiedampfungsfaktoren von 2 langs der gesamten Lange des Profils waren tj^pisoh. Dies stimmt grBenordnungsmaBig berein mit Dampfungsraten, wie sie infolge von Bodem'eibung erwartet werden. Jedoch ^vurde die starke Gezeitemnodulation, die von der Theorie fr ein quadratisches Bodenreibungsgesetz vorhergesagt wird, nicht beobachtet. Gegenwinde batten keinen EinuB auf die Zerfallsrate. Durch Berechnungen lassen sich auch Wellen-Wellen-Wechselwirkimgen oder turbulente Dissipation auBerhalb der Bodengrenzscliicht als wirksame Mechanismen der Dnungsdampfmig ausschlieBen. Wir folgern daraus, daB entweder das allgemein akzeptierte Reibungsgesetz wesentlich geandert werden muB oder daB ein anderer Mechanismus, wie z.B. Streuimg durch BodenunregelmaBigkeiten, die Ursache der Dampfung ist. Die Dispersionscharakteristiken der Dnmigen deuteten auf recht nahe gelegene Ursprmigsgebiete bin, fr die das klassische (5-Ereignis-Modell im allgemeinen nicht anwendbar war. Eine starke Doppler-Modulation durch Gezeitenstrme wurde beobachtet.
Jlesures de la croissance des vagues dues au vent et du comblement de la houle dans Ie Projet commuu relatit aux vagues do la mer du Nord (Joint North Sea Wave Project - JONSWAP) (Rsum).
On a mesiu' les spectres des vagues, Ie long d'ime coupe s'tendaut sm- 160 kilomtres, dans la mer du Nord , l'Ouest de Sylt, pendant une priode de dix semaines en 1968 et 1969. Au corn's de l'exprience principale, en juillet 1969, treize stations d'observation des vagues furent mises en service, dont six continurent les mesures pendant les deux premires semaines d'aot. On a fait une experience pilote plus rduite en septembre 1968. On a galement mesiu' les comants, les mares, les differences de tempratm-e entre Fair et la mer ainsi que la tm'bulence dans la couche atmosphrique voisine de la surface. Le hut de l'exprience (indic[u dans la premire partie) tait de dterminer la structure de la fonction-origine determinant l'quilibre de l'nergie du spectre de la vague, en insistant particulirement sur la croissance de la vague quand le vent souffle de terre de fajon constante (deuxime partie) et sur l'attnuation de la houle par fonds de profondeur limite (troisime partie). Les fonctions-origines du spectre de la vague souleve par des vents de terre ont une allure plus ou moins caractristique, en rapport avec le dcalage du sommet escarp du spectre vers les frquences plus basses. La distribution de la fonction-origine selon deux lobes peut s'expliquer quantativement par le transfert non linahe du aux actions rciproques vague-vague, en rsonance (Dispersion de Bragg du second ordre). L'volution d'un sommet accentu et son dcalage vers les frciuences plus basses peuvent aussi tre interprts comme mie caractristique de stabilisation de ce processus. Pom' les petits dplacements, l'quilibre de l'nergie se realise principalement entre celle qu'apporte le vent dans la region centrale du spectre et le transfert non lineaire d'nergie loin de cette region: - Vers les vagues courtes, o elle se dissipe, ^ et vers les vagues plus tongues. La plus grande partie de la croissance de la vague, sur la face avant du spectre, peut tre attribue au transfert non lineaire vers les vagues plus longues. Pom' les dplacements courts, environ (80 + 20) % de la force vive transfre k travers l'interface air-mer entre dans le champ de la vague, d'aprs les mesures d-ectes de Dobson, relatives au travail que les pressions extrieures effectuent sm' les vagues. Environ 80 a 90 % du flux de force cre par la vague passe dans les coiu'ants par la voie d'un transfert non lineaire dans les vagues ooiu'tes et d'une dissipation ultrieure, le reste demeui'e dans le champ de la vague et se trouve emport au loin. Pour les dplacements plus
Hasselmann et al., JONSWAP
amples, l'interprtation de Tquilibre de l'nergie devient plus incertaine car on ne sait rien de la dissipation dans la rgion des basses frquences du spectre. Une dissipation nuUe dans oette gamme de frquences donne un flux minimal de force vive atmosphrique dans le champ de la vague, de l'ordre de 2 0 ^ 10)'^ " ' ^^^ transfert total de force vive , travers l'interface air-mer, mais des rapports allant jusqu'a 1 0 0 % sont concevables si la dissipation est importante. En gnral, les rapports (dcoulant du transfert non lineaire d'nergie) restent en dedans de ces limites dans mie large gamme (cinq dcades) de dplacements cernant Fenvironnement de la vague et, aussi, les donnes prsentes du champ. Cela laisse penser que les chelles du spectre s'ajustent d'une fagon continue, de sorte que les actions rciproques des vagues quilibrent exactement l'nergie apporte par le vent. On peut expliquer ainsi, entre autre particularits, la dcroissance observe de la constante de P h i l l i p s avec le dplacement. Les taux de comblement dtermins pom la houle arrivant sm des fonds limits variaient normment mais les facteurs d'attenuation des deux nergies siu- la longueur de la coupe taient caractristiques. Leur ordre de grandeur correspond aux taux d'amortissement escompts pour le frottement sm' le fond. Toutefois, on n'a pas observ la forte modulation due k la mare, thoriquement prvue dans le cas d'une loi quadratique du frottement sur le fond. Les vents contrahes n'ont pas affect le taux de comblement. Les calculs liminent aussi, comme moanismes rels de l'attnuation de la houle: - Les actions reciproques vagues-vagues, - ou la dperdition due . la turbulence hors de la couche voisine du fond. Nous oonclm'oiis, ou bien c[u'il faut modifier profondment la loi gnralement admise pour le frottement, ou que quelqu'autre mcanisme, telle la dispersion cause par les irrgularits du fond, est la cause de l'attnuation. Les caractristiques de dispersion des boules indiquent plutt des origines voisines, pour lesquelles le modle classique du cas tait gnralement inapplicable. On a galement oliserv ime forte modulation D o p p l e r due aux courants de mare.
1. The Experiment 1.1. Intioductiou The last fifteen jêars have seen considerable progress i n the theory of \\dnd-generated ocean waves. These advances have been paralleled hj the development of improved mathematical techniques and computer facilities for numerical wave forecasting. However, critical tests of theoretical concepts and numerical forecasting methods have been severely limited i n the past by the lack of detailed field studies of wave growth and decay. Thus considerable effort has been devoted to the theoretical anatysis of many of the processes which could affect the spectral energy balance of the wave field - the generation of waves by wind (bj^ a variety of mechanisms), the nonlinear energj' transfer due to resonant wave-wave interactions, the energy loss due to white capping, interactions between very short waves and longer waves, and the dissipation of waves i n shallow water by bottom friction, to list some of the more important contributions - but only very few Lave been subjected to quantitative field tests. Even where field experiments have been specifically designed to study these questions, limitations i n available instrumentation have often led to incomplete coverage of the wave field, thus handicapping the interpretation of the data. Consequently, present wave prediction methods have necessarily been based on rather subjective conjecttu'es as to the form of the som'ce function governing the fundamental energy balance ec(uation. The Joint North Sea Wave Project (JONSWAP) was conceived as a cooperative ventm-e by a number of scientists i n England, Holland, the United States and Germany to obtain wave spectral data of sufficient extent and densit}^ to determhie the strnctiue of the soiu-ce function empu'ically. I t was decided that an arraj' of sensors operating c[uasi-continuouslj' for a period of several weeks was the most straightforward method of obtaining adecjuate sampling clensitj' of the wave field with respect to frec[uenc3', propagation direction, space and time, under the deshed wide variety of external geophysical conditions. The array that was eventually used consisted of thirteen wave stations spaced along a 160 km profile extending westward from the Island of Sylt (North Germany) into the North Sea. A l l instruments provided the usual frec[uency power spectra, while six of the stations (5 pitch-roll buoj^s and 1 six-element array) yielded directional information i n addition. The complete profile was i n operation for four weeks during July 1969; another six weeks of data were obtained fi'om a smaller number of stations during the two weeks immediately following the main experiment and i n a pilot experiment i n September 1968. The enthe set of time series was spectral anatyzed, but the interpretation of the data is restricted i n the present papers to selected cases falling into either of two partioularlj' simple categories: the evolution of the spectrum under stationary, offshore wind conditions (Part 2), and the attenuation of incommg swell (Part 3). Over 2000 wave spectra were measm-ed; about 300 of these represented reasonably well-defined c|uasi-stationary generation cases, of which 121 con-esponded to "ideal" stationary and homogeneous 'wind conditions; 654 spectra contained well-developed peaks suitable for the swell attenuation study. Wave Growth The present wave generation study may be regarded as a natural extension of the pioneering work of R. L . S n y d e r and C. S. Cox [1966] and T. P. B a r n e t t and J. C. W i l k e r son [1967]. Both of these experiments utilized a single mobile wave station to determine the evolution of the wave spectrum i n space and time.
_ S n j r d e r and Cox measured the growth rate of a single frequeney component (0.3 Hz) by towmg an array of wave recorders downwind at the group velocity of the component. They found exponential growth i n accordance m t h the linear instability mechanisms proposed by J. J e f f r e y s [1924] and J. W. M i l e s [1957], but at a rate almost an order of magnitude greater than predicted by either theory - or by the later instability theories which took more detailed account of the turbulent response characteristics of the atmospheric boundary layer g.g. C M . P h i l l i p s [1966], K . H a s s e l m a n n [1967, 1968], R. E. D a v i s [1969, 1970], P. L o n g [1971], D. E. H a s s e l m a n n [1971]). Moreover, assuming that the observed growth rates were due enthely to a linear input from the atmosphere, S n y d e r and Cox inferred a niomentum tran.sfer from the atmosphere to the wave field which was several times greater than the k n o m i total momentum loss from the atmosphere to the ocean. As shown below, it appears that this paradox can be largely resolved by taking account of the nonlinear wave interactions, which we found to be responsible for the major part of the observed wave growth on the forward face of the spectrum. The evolution of the complete (one-dunensional) wavenumber spectrum under fetohlimited conditions was fii'st measured by T. P. B a r n e t t and J. C. W i l k e r s o n [1967] using an au-borne radar ahuneter*. Only two wave profiles were AoMm, but again exponential growth rates were found i n general agreement Awth S n y d e r and Cox's results. I n addition, the measurement of the complete spectrum revealed a significant overshoot effect; the energy at the peak of the spectrum was found to be consistently higher by factors between 1.2 and 2 than the asymptotic equilibrium level approached by the same frequency at large fetches A similar effect was observed by A. J. S u t h e r l a n d [1968] and by H . M i t s u y a s u [1968a, 1969] m a wind-wave tank (see also T. P. B a r n e t t and A . J . S u t h e r l a n d [1968]) A l l authors speculated that the effect may be due to nonlmear processes. This interpretation found fm-ther support through H . M i t s u y a s u ' s [1968b] wave taifii mea.surements of the decay of a random wave field, which could be explained quite well by T. P. B a r n e t t ' s [1968] parametrical estmiates of the nonlinear energy transfer due to resonant wave interactions ( l i ^ H a s s e l m a n n [1962, 1963a, b]). (In then Pacific swell study, P. E. Snodgrass G. W. Groves, K . E. H a s s e l m a n n et al. [1966] had sunilarly concluded that the decay ol new swell close to a generation area was i n accordance with this process.) Useful iifformation on wave growth can also be obtained indirectly from measurements at a single fixed location under different wind conditions. Using dimensional arguments S A. K i t a i g o r o d s k i i [1962] has suggested that for a stationary, homogeneous wind field blowing orthogonally off a straight shore, the appropriately nondimensionalized wave spectrum shoid be a universal function of the nondimensional fetch x = gxjul only where x IS the (dimensional) fetch, g is the acceleration of gravity and is the friction velocity Thus wave measurements obtained at a fixed location for different wind speeds can be translated into an eqmvalent dependence on fetch at a fixed -\vind speed. H . M i t s u y a s u [1968a, 1969], H . M i t s u y a s u , R. N a k a y a m a and T. K o m o r i [1971], P. C. L i u [1971] and others have applied K i t a i g o r o d s k i i ' s scaling law successfully to relate wave tank and field measm-ements and characterize the basic properties of fetch-limited spectra. The JONSWAP data is i n general consistent with K i t a i g o r o d s k i i ' s scaling hypothesis and confirms many of the spectral featui-es summarized by these authors: pronounced overshoot factors, narrow peaks (considerably sharper than the fully-developed P i e r s o n - M o s k o w i t z [1964] spectrum), very steep forward faces (with associated rapid growth rates of wave components on the face), and an /-^ high frequency range with a fetchdependent 0. M . P h i l l i p s ' [1958] "constant". A principal conclusion of om- wave growth study is that most of these properties can be explained quantitatively or qualitatively by the nonlinear energy transfer due to resonant wave-wave interactions (second order Bragg scattering). Thus the evolution of a sharp spectral \T T * ^"îl'""!' experiments using a laser altimeter have been repeated recently by D. B. Ross, De L e o r i b u s l i m ] Conaway [1970] and J. R. Scbule, L . S. Simpson and P. sl
12
Erganzungsheft zui- Deutschen Hydi-ographischen Zeitscln-ift. Reihe A (8), Nr. 12, 1973
peak is found to be a self-stabilizing feature of this process. The continual shift of the peak towards lower frequencies is also caused by the nonlinear energy transfer, and explains the large growth rates observed for waves on the forward face of the spectrum. Por small fetches, the energy balance i n the main part of the spectrum is governed by the energy input from the atmosphere, the nonlinear transfer to lower and higher frec[uenoies, and advection, dissipative processes apparently playing only a minor role. This balance determines the energy level of the speotnim, i.e. the value of P h i l l i p s ' constant, and the rate at which the peak shifts towards lower frequencies. Assuming a friction coefficient cô of the order of 1.0 x 10-^ this pictm-e implies that nearly all of the momentum lost from the atmosphere must be entering the wave field - i n agreement ^^dth P. W. Dobson's [1971] correlation measm-ements of wave height and surface pressure. About 10 % of the momentum fiux from the atmosphere to the wave field is transfen-ed to lower-fr-equency waves, where i t is advected away. The rest is transferred to short waves, where i t is converted to cm-rent momentum by dissipative processes. W i t h increasing fetch, the wave-wave momentum transfer to high wavenumbers decreases, representing only 13 % of the total momentum fiux across the ah-sea interface in the Ihnit of a fully-developed P i e r s o n - M o s k o w d t z spectru,m. The net momentum fiux T f r o m the atmosphere to the wave field at large fetches depends on the unknown dissipation i n the main part of the spectrum. Zero dissipation i n this range implies a lower limit for T,^ of about 20 % of the total ah-sea momentum flux - but this value could also lie closer to 100 %, as i n the small-fetch case, i f dissipation is an important factor i n the energy balance of the principal wind-sea components. Although the present study has helped to clarify the general struotm-e of the energy balance of fetch-limited wave spectra, a number of questions remain. The mechanism by which energy and momentum is transferred from the wind to the waves could not be investigated. Measm-ements with hot-wu-e instruments, wind vanes and cup anemometers yielded atmospheric tm-btence spectra and the net momentum flux fr-om the atmosphere to the ocean, but the instruments were too far above the surface to resolve the interactions between the atmospheric boundary layer and the wave field. A second open problem concerns the dissipative processes balancing the nonlinear energy flux to short waves. This question is of considerable interest not only w i t h regard to the wind-wave energy balance, btit also for the application of microwave techniques to the measurement of sea state fr'om sateUites (cf. K . H a s s e l m a n n [1972]). Finally, the transition from fetch-limited spectra to a f u l l y developed eqiulibrium spectrum poses a number of questions which could not be investigated within the finite range of fetches available i n this study. Swell A t t e n u a t i o n For the generation studies, the water depth along the entire profile was sufficiently deep to be regarded effectively as infinite. However, much longer incoming swell components normally "felt bottom" along most of the profile, thus affording a good opportunity to investigate the effects of bottom topography on swell propagation. I t is well known that bottom effects play a significant role i n the energy balance of waves i n relatively shoal areas such as the North Sea or other parts of the continental margins. Most investigators have attributed the observed modifications i n the form of the f u l l y developed spectrum (e.g. J. D a r b y s h i r e [1963]) and the enhanced damping of the swell to bottom friction. Assuming a cûadratic friction law, C. L . B r e t s o h n e i d e r and R. 0. R e i d [1954] estimate a friction coefficient of the order 10"^. Similar values were found by K . H a s s e l m a n n and J. I . C o l l i n s [1968], who applied a more detailed spectral formulation of the bottom fr-iction theory to hm-ricane swell data i n the Gulf of Mexico. The considerably larger decay rates found by H . W a l d e n and H . J. R u b a c h [1967] i n the North Sea are also consistent with the H a s s e l m a n n and C o l l i n s theory i f the tidal cm-rents are taken into account. However, the extensive swell data available through this experiment, while i n order of magnitude agreement with previous observations of swell attenuation rates, contradict a
Hasselmann et al., J O N S W A P
Erganzungsheft zur Deutscfien Hydrograptiischen Zeitschrift. Reihe A (8), Nr. 12,1973
15
significant prediction of the theory. A quadratic (or any nonlinear) fi'iction law should lead to a strong modulation of the swell decay rates by tidal currents. Experimentally, no such variation is foiuid. We suggest two possible explanations - neither of which we have been able to test i n the present experiment. I t is conceivable that the quachatic friction law, although shown to be an acceptable fii'st-order approximation for sinusoidal waves i n a wave tank (J. A. P u t n a m and J. W . J o h n s o n [1949], P . P . Savage [1953], K . K a j i u r a [1964], Y . I w a g a k i , Y . T s u c h i y a and M. S a k a i [1965], I . G. J o n s s o n [1965]), is not valid for the time-dependent boundary layer generated by the superposition of a random wave field on a tidal current i n the real ocean. A n objection to this explanation is that a complete independence of the swell attenuation on the tidal cuiTents implies a linear friction law, which seems improbable for a tm'bulent boundary layer. Alternatively, the observed swell decay may be due to processes other than bottom friction. Estimates of the damping due to interactions ^vith turbulence distributed throughout the water column (K. H a s s e l m a n n [1968]) indicate that this process is too small by several orders of magnitude. A more promising possibility is the baokscattering caused by u-regularities of the bottom topography at scales comparable with the wavelength of the swell. Recent computations by R. L o n g [1972] show that this interaction could account for the observed attenuation rates assuming r.m.s. bottom hregularities i n the appropriate wavelength range of the order of 20 om. Experimentally, this process should be identifiable by the presence of offshore-propagating swell components whose energy increases with distance from shore (in contrast to the decreasing energy of swell which may be refieoted fi-om the shore). Unfortunately, the dh'eotional resolution i n the present experiment was insufficient to resolve secondary swell components propagating i n a dhection opposite to the main swell, 1 .2. The Wave Profile The site of the wave study and the 160 k m long measming profile are shown i n Figures 1.1 and 1.2. The area was chosen on account of its relatively smooth bottom topography, moderate tidal cm-rents and convenient logistics. The main experiment ran from July 1 until July 31, 1969, but measurements continued from August 1 to August 15, 1969 along a reduced profile (Stations 1-5 and 8). Diu-ing these six weeks, 30-minute wave recordings were obtained contmuously at 2 or 4 hom-ly intervals. Fm-ther measm-ements were also made along a 10-station profile dm-ing a pot experhnent from September 1 to September 30, 1968. I n this earlier experiment measm-ements were limited to "hiteresting" cases of wave generation or incoming swell, and the times of recordmg at different stations were staggered to allow for the propagation time of the expected prmeipal wave components. This mode of operation proved to be too demanding on the reliabUity of weather forecasts and commmiications, so that i n the main experiment the following year a more rigid, predetermined schedule of recordings was introduced, i n which the only on-spot decisions involved changes from standard four-hourly recording to two-hourly measm-ements dm-ing east-wind generation conditions. The depth distribution along the profile lay i n a convement range enabling the study of both the generation of waves dm-ing offshore winds and the attenuation of long swells rmunng inshore. For short, fetch-limited waves generated by local east muds, the depth could be regarded as essentiaUy infinite along most of the profile, cf. Figm-e 1.3. The om-ves indicate the depth H = at which the principal wind-sea components would begin to be = ^ for a given wind speed
infiuenced appreciably by the bottom; the wavelength
and fetch was evaluated from the appropriate empu-ical fetch dependence of the frecjuency of the spectral peak. Fig. 2.6. The bottom influence is seen to be negligible for all wind speeds relevant for this study (the maximum east winds occiu-ring dm-ing the experiment were of the order of 15 meter per second). On the other hand, a typical swell component of
16
ErganzLingsheft ziu' Deutschen Hj^ch-ographischen Zeitschrift. Reihe A (8), Nr. 12, 1973
10 seconds period, corresponding to a deep water wavelength of about 150 ni, is appreciably influenced by the bottom along the enthe profile, so that bottom friction or bottom scattering effects shoidd be easily observable. A variety of wave instruments were used i n the experiment (Fig. 1.2). A t Stations 1, 2, 3, 5 and 8 (we refer here and i n the following to the July 1969 profile) the wave motion was measm-ed by a small fioat moving within a perforated pipe mounted on a post. A t Station 4 a linear dhectional array of six wave recorders was installed at variable separations along a line orthogonal to the profile (of. Section 1.5). The wave sensors consisted of sub.sm'face pressm-e transducers mounted on posts, with the exception of a closely spaced pah of resistance wave staffs used to resolve the short wave components. Tethered "waverider" buoys, which converted measm-ements of vertical acceleration internally to wave height, wêre used at Stations 7, 9 and 10. A l l of these nearshore instruments were unattended and/or remotely controlled and generated F M output signals which were telemetered to a shore station at the base of the profile. This technique was not readily applicable to the more distant Stations 11-13 and Station 6, which lay within a shippmg lane. These stations were occupied by pitch-roll buoys deployed from ships. The pitch-roll buoy measm-es the vertical acceleration and both components of wave slope, thus giving a dn-ectional resolution of the wave field roughly comparable to the array at Station 4 (cf. M . S. L o n g u e t - H i g g i n s , D. E. C a r t w r i g h t and N . D. S m i t h [1963]). A l l instruments had high-frec[uency cut-offs between 0.7 and 1 Hz, corresponding to wavelengths between 3 and 1.5 m. Digitizmg rates v.'ere normally 2 per second, ecpuvalent to 1 Hz Nyquist frequency. For iiitercalibration pm-poses Station 10 was occupied dm-ing the first half of the 1969 experiment by two instruments, a waverider and a pitch-roll huoy. This also provided a redundancy safeguard, and was useful i n occasionally freeing the ship at station 10 to assist in servicing the unattended stations. I n the second half of the experiment the pitch-roll buoy of Station 10 was moved to Station 7 and the waverider of that station was switched to Station 5 for comparison with the wave fioat instrument. I n addition, a number of intercomparison nms between various combinations of instraments were made during the 1968 experiment. A continual check on calibration consistency was fm'thermore provided by the intermeshed distribution of different types of instruments along the profile (cf. Section 1.4). 1.3. Other Field Measm-ements To define the envhonmental parameters relevant for the energy balance of the wave field, -winds, atmospheric momentum fiuxes, cmrents, tidal heights and afr-sea temperatm-es were monitored at various stations along the jirofile. Wind speeds and dfrections and tidal heights were recorded contmuously at Stations 2 and 8. Winds were also measmed concm-rentty with ah-sea temperatm-es every two hours on the ships. A t Station 8, atmospheric turbulence was measm-ed at two levels, 5 and 8 m above mean sea level. The sj'stem at each level consisted of a combmation of standard hot-whe instruments, together vdth cup anemometers and vanes intended for calibration purposes (K. E n k e [1973]). A l l tm-bience data were telemetered to shore together with the concurrently taken wave data (of. Section 2.3). Contmuous cm-rent recorduigs were made throughout the 1968 and 1969 experiments at a number of depths and stations (Fig. 1.3). Emphasis on current measurements was based on the theoretical anticipation (K. H a s s e l m a n n and J. I . C o l l i n s [1968]) that eim-ents near the bottom would have a strong influence on the attenuation of long swell by bottom friction. I t was also conceivable that vertical cmTent shear would influence wave generation. However, no detectable current influence was found i n either the generation or the swell attenuation study. Since i n most cases the observed ciUTents did not deviate appreciably from the tabulated tidal cm-rents (particularly the near-bottom cm-rents) i t proved adequate
Hasselmann et al, J O N S W A P
1 7
0.40
ST. 5 JULY 2 2 1 9 6 9 2 0 0 0 X WAVE R I D E R + WAVE F L O A T
ST. 5 JULY 2 2 1 9 6 9 1 6 0 0 X WAVE R I D E R + WAVE F L O A T
1.00
ST. 4 JULY 2 2 1 9 6 9 1 8 3 0 X PITCH ROLL WIRE
080
ST. 4 JULY 2 2 1 9 6 9 1 6 3 0 X PITCH ROLL WIRE
+ RESISTANCE
-t- R E S I S T A N C E 060
0.40
O20
0.6
0.00 0.0
FREQUENCY
(Hz)
Fig. 1.4 a-d. Intercomparison measurements of frequency wave spectra
POWER SPECTRUM
(rrt/Hzj
POWER SPECTRUM (m^/Hz)
o o 1
l"-^ )
PITCH - ROLL/
St. 5 Sept 6 19681230
'
/ " A
\l
m - /'
19
for the negative ontcome of the eorrelation analysis described i n Parts 2 and 3 to worli with tlie tabulated rather than the observed currents. Independent of the wave project, however, the measui'ement of oun-ents was also motivated by interest i n the baroclmio response of the sea to a variable wind field. Por this pm-pose a continuous hydro-station program was carried through from the ships at the outermost stations 11, 12, and 13. This work vdll be reported elsewhere (G. B e c k e r , K . P. K o l t e r m a n n , G. P r a h m et al. [in prep.]. 1.4. Intercomparison of Frequency Wave Spectra The choice of a variety of wave measmdng technic[ues was necessitated by the limitations of V H F and U H F telemetry ranges, the depths at which wave posts could be economically installed, and i n part by the prior experience of participating groups. A prime concern of the experiment was therefore to establish the compatibility of the different technic[ues used. Figs. 1.4a-h show a series of one-dimensional frequency spectra measm-ed at the same location and time by pahs of different instruments. I n general, the agreement is excellent. Not all combinations of different instrument pahs were compared, but an independent conthmal check on the iiitercalibration consistency of all instruments was provided by the dovetailed distribution of different instrument types along the profile: at least one instrument of each group was fianked at some point i n the profile by instruments of another type. I n all cases studied, the spectral distributions showed a smooth variation with station number, without discermble discontinuities at the instrument transition points (with the exception of one generation case on September 11, 1968, i n which thi'ee of the runs indicated an - mrresolved systematic error i n the pitch-roll buoy measm'ements at Station 6, which were then discarded). Although coiitrarj' to standard experimental practice, the simultaneous use of a number of independent measmement teohiiic[ues proved to be an asset i n establishing a broad calibration base and providing continual calibration monitorhig thi-oughout the experiment. The restriction to a smgie type of instrument has advantages i n linear problems involving the comparison of energy levels at the same frequency. I n this case the common instrumental transfer function can be eliminated. However, i n the present experhnent the evolution of the wa.ve spectrum was fomid to be strongly dependent on noifiinear processes for which the determination of absolute energy levels i n different frecjfuency bands was essential. I t appears that Avith very few excejitions the one-dimensional wave spectra measured by all instruments were acom-ate to ^vithin the attauiable statistical resolution. I n a few cases the records were contaminated by noise, most frequently due to radio interference i n the telemetry links. The noise spectrum was generally white, and i f sufficiently low was simply .subtracted. For noise levels greater than 20 or 30 % of the total energy, the measm'ement was discarded. About 90 % of the data yielded useable spectra i n this sense. I n the case of the swell studj' (Part 3), the figm-e was somewhat reduced by the inherent low-frequency noise of the pitch-roll buoys, which sometimes masked low-energy swell peaks of long period. 1.5. Dhectional Measm'ements The dh-ectional distribution of the wave field was determined from the pitch-roll buoys (Stations 6, 10, 11, 12 and 13) and the linear array at Station 4. Both instruments yielded oifij^ a limited number of moments of the energy distribution (, 0) with respect to propagation dhection 9 for a fixed frec(uenoy . The moments obtained from the two types of instrument differ, so that an intercomparison is, i n general, possible only i n terms of a given parametrical model of the angidar distribution function. The dh-ectional characteristics of the pitch-roll buoy (hi its earliest instrumental form) are described by M. S. L o n g u e t - H i g g i n s et al. [1963] and by D. E. C a r t w r i g h t and N . D . S m i t h [1964]. Essentially, the instrument records three channels, two components (1) and (2) of the wave slope, and the vertical acceleration (3). A fourth (compass) channel enables (1) and (2) to be re-oriented to geographical axes. The auto-spectrum of (3), Ogg, is
20
Erganzungsheft zur Deutschen Hych'ographisehen Zeitschrift. Reihe A (8), Nr. 12, 1973
easily translated into the total energy spectrum of vertical displacement, E{f), and the autospeotra of (1) and (2) combine to give a dhect measm-e of total wavenumber k, through the relation
Three of the cross-spectral elements C,j., Qij, have zero expectancy, while the remaining three, together m t h Oj^ Câ, provide four explicit and independent angular moments of the dhectional spectrum, of the form
]" F ( f , e ) ( k e y d e ,
71
n=l,2.
Eor reasons given below we i n fact found use only for moments of order n = i, and these may be expressed as moments of the normalised angular distribution s{f, ), where ! ' ( / , 0) = s (/, 0) E (/), Avith Y
It
s (/, 0) d0 = 1; thus
where
(/cj, k^) = (k cos 0, k sin 0), and k; = J kjSdO. n I n the case of the linear array, the wave heights d and l^j measured at any two locations spaced at a distance Vjj normal to the profile yield dhectional moments
where the angle of wave propagation 0 is measm-ed relative to the profile dhection. The array consisted of a superposition of two 4-element arrays differing by a factor of 4 i n scale (Fig. 1.5). The coordinates of each 4-element array were chosen as Xj = 0, D,
L I N E A R ARRAY,STATION
z
1-
(1J
1 ' " 1
'
* 3D'
2D' 0 WAVE STAFF e PRESSURE TRANSDUCER D=7m D'=4D = 28m
1, 2D 1 ^ .1. to O ^ t)
3D
IDI,
.1
Fig. 1.5. Tlie linear array at Station 4 4D, GD {D = 7 m and 28 m) yielding the 6 lags r^j = D, 2D, .. . 6D. Each subarray provided dhectional information (depending somewhat on the distribution assumed) i n wavelength bands between about 10 m and 150 m, and 40 m and 600 m, respectively. On account of theh mhTor sj^mmetry linear arrays are unable to distinguish between waves approaching the array at the same angle but from opposite sides of the array, i.e. at
Hasselmann et al., JONS^VAP
21
anoies 9 and 70 - 0. This ambiguity was not regarded as a serious limhation i n the present experiment. The smaller-scale array was designed to resolve the dhectional distribtition of short locally generated weaves imder east-wind condhions, whereas the larger-scale array was iised to determine the propagation dhection of incoming swell. I n both cases only one of the two waveimmber half-planes contained signhicant energy. A number of techniques have been developed for the dhectional analysis of muhisensor arrays I n our case, the details of the dhectional distributions tm-ned out to be unerhical for the interpretation of the data. Thus i n the wave-growth study, values of the mean propagation dhection and mean square dhectional spread proved adequate to compute and interpret the somce function to wdthin the error limits imposed by other factors of the experiraent The same parameters were also sufficient to characterize an incoming swell beam. (Spectra containmg two swell components with the same frequeney but different propagation dhections were very seldom.) n . .1 .î Consecpiently, the dhectional data were analyzed and compared prunariiy with respect to the two simplest dhectional parameters, defined (in terms of the phch-roll data) as 0 = arotan^ ^]''s{9)9d9,
J 1^
s{9){9-9j'd9
I n the I h n h of a narrow beam, these are identical with the mean dhection and r.m.s. dhectional spread, as defined by the integrals on the right hand side*. Although this made oifiy partial use of the avafiable dhectional resolution, no attempt was made at a systematic, high-resolution analysis, except for occasional spot checks of the consistency of the remahihig dhectional moments with respect to the two-parameter dhectional Apart f r o m the insensitivity of om- conclusions to the details of the dhectional distribution, this approach was also motivated by experimental difficuhies. Although extensively tested i n individual measurements, the compass units of the pitch-roll buoys were apparently unable to withstand the continual mode of operation requhed for this experhnent, and only one of the five instruments yielded rehable dhectional data throughout the enthe period. Also an early failm-e of the dighal data logging system for the linear array meant that much of this data had to be recovered from analog traces by pamstaking semiautomatic digitalization. The restriction to two dhectional parameters had the advantage i n this case that they could be derived from only two sensors of the array. Analogous variables to 0 and 0, for two wave sensors separated by a distance r orthogonal to the profile may be defined as /arctan QIG 9' = arcsin'
kr
i - R ' -, {kr cos 0 ; ) ^
1 .1 1 where the coherence
J, M=
f(^'+Q
and we have dropped the subscripts ij on C, Q and r. For a narrow beam 9'^, 9', - > 0 ^ , 0^
* I n contrast to the definitions in terms of kj, the latter integrals represent meaningful quantities only m the narrow beam Irniit. On acooimt of the non-periodicity of the "^tegrand, they are not invariant with respect to redefinition of the 0 interval, say from (-Tt, n) to (, Zn), for arbitrary distributions s (0).
22
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8"), Nr. 12, 1973
To compare the dhectional estimates from the pitch-roll and linear array measm-ements, a standard spreading function model was assumed,
where / j is a normalismg parameter, making

+ 11
s d = 1.
71
Pig. 1.6 a shows the relation between 9^ and 9'^ for hr = 1 and different beam widths 9^, as characterized by the parameter p. The corresponding relation between 9^ and 9' as a function of the propagation dhection 9^ is shown i n Eig. 1.6b. Shmlar curves were computed for other models of the spreading function, some including a skewness factor. The transformation was fomid to be rather insensitive to the choice of model. W i t h the aid of Figs. 1.6 a and b, the linear-array dh-ectional data could be converted to equivalent pitch-roll dhectional parameters. A typical intercomparison run at Station 4, shoAving satisfactory agreement of the dh-ectional parameters after conversion, is presented i n Fig. 1.7. 1.6. Cases Studied Fig. 1.8 shows the general wind history for the three periods of the experiment. Wind dhections are defined relative to the proffie, arrows pointing vertically upwards representing offshore winds parallel to the jirofile. The arrows represent averages of the winds measm-ed at given time along the profile. Wave groivth was investigated for the periods marked G; the symbols I G refer to "ideal generation" conditions of high stationarity and homogeneity, which were selected for more detailed study. A more complete time history of the vnnd fields during these periods and weather maps are presented i n Part 2. Periods marked S were selected for the swell attenuation study. The term swell is normally applied to waves no longer being actively generated by the wind, i.e. i n the present case, waves travelling inshore at phase speeds exceeding the onshore component of the local wind velocity. Usually, these conditions apply only after the waves have traveUed a sufficient distance fi-om the wave som-ce for dispersion to have converted the swell spectrum into the classical form of a narrow low frequency peak. However, i n the present experiment several cases also occurred w^here the local -wind field died down rather rapidly, leaving a locally generated, broad-band swell spectrum. We have investigated here only "classical" swell cases i n which the width of the swell peak was narrow compared with the swell frecjuenoy. This was because om- main interest concerned the bottom interactions, which we wished to separate as far as possible from other dynamical processes. Computations (K. H a s s e l m a n n [1963b]) and experiments (F. E. Snodgrass et al. [1966], H . M i t s u y a s u [1968b]) indicate that noifiinear wave-wave interactions are important for the energy balance of a broadband, newly created swell specti-um, whereas they can be neglected for a well dispersed swell train* (pro-sdded the frequency lies below the frequencies of the local background wind sea). The generation and swell cases studied i n this paper have been chosen from a much larger ensemble of measured spectra. They represent the simplest situations for investigating the basic processes that presumably also control the evolution of the wave spectrum in the general case of an arbitrary vnnd field and boundary condition. The interpretation of the * Tlie essential condition is that the swell energy is small relative to a newly created swell. I n sorne of the quasi-stationary swell cases considered in Part 2, this was caused by angular dispersion alone, rather than the usual combination of angular and frequency dispersion.
23
es (PITCH - ROLL BUOY) Fig. 1.6. a,b. Relation between parameters ctiaracterizing ttie mean direction (a) and directional spread (b) for ttie pitcti-roll buoys and the linear array. A cosê spreading factor is assumed
24 Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973
Fig. 1.8. Wind history during Sept. 1968 and Julyt - Aug.15,1969. Wind arrows represent averages along the profile Offshore winds parallel to the profile point upwards
25
remaining data would require more sophisticated analysis technic[ues than used here, includmg the numerical integration of the energy balance ec[uation for time and space dependent v/ind fields. Edited data tapes of the complete set of wave spectra are available on recpiest from the National Oceanographio Data Center, U.S.A., the Institut f r Geophysik, University of Hamburg, or the Deutsches Hydrographisches Institut, Hamburg. Wind and other envhonmental data are available from the latter two institutions. 1.7. Logistics Overall logistic coordination of the experiment was carried out by the Deutsches Hydrographisohe Institut (HW)*, Hamburg, with assistance from the Institut f i h Geophysik, Umversity of Hambirrg (I^H). Site preparation and technical coordination of field operations were handled by A. Hechich through the Deutsches Hydrographisches Institut. Scientific communication dmdng the experhnent was maintained from the shore-based telemetry and radio station (1968: K H ; 1969: H W , K H ) for which accomodation was kindly provided by U . Jessel, Institut f i h Bioklimatologie und Meeresheilkmade. Withhi this general logistic framework, each participatmg group canded individual responsibility for its contribution to the experhnent, as follows (see also T. P. B a r n e t t [1970]): Deutsches Hydrographisches Institut. 1968: Wave posts at Stations 2 and 8 (Carstens). 3 phch-roll buoys operated from "Walter K r t e " (KR), " 0 T 2 " (HC, H . Erentz), and "Gauss" (H. J. Rubach, P. Lmdner). Tide gauges at Stations 2 and 8. Cm-rent meters. 1969: Wave posts at Stations 1, 2, 3, 5 and 8 (U. Carstens, Rossfeld, H G ) ; 4 phch-roll buoys operated from "Waher K r t e " (KR), " 0 T 2 " (HC, H . Erentz), "Gauss" ( H . J . Rubach, P. Lindner), and "Holnis" (R. Pirhrmann); (one of the phch-roll buoys wâs loaned from the National Institute of Oceanography). The telemetry and data acquisition system for the wave posts and turbience instmmentation were b u t by U . Carstens. Wave posts for the D H I wave instraments were built and installed (by jetting into the sandy bottom) by Harms & Co. (S. Knabe). Tide gauges at Stations 2 and 8. Institut fih- Geopliysik, Hambm-g. 1969: Tm-bience measurements, Station 8 (O. Kertelhem, K E , DO). The turbulence instramentation was b u t by 0. KerteUiem. Konmklijk Nederlands Jleteorologisch Instituut. 1968: Thi-ee waverider buoys (AM). 1969: Waverider buoys at Stations 7, 9 and 10 (EB). The waverider system, complete -with telemetry and data acqirisition system, was built by Datawell NV. National Insthute of Oceanography. 1968: Two phch-roll buoys, operated from HMS "Enterprise" (JE, D. Bishop) and HMS "BuUdog" (N. Smith, P. Cofiins). 1969: Pitch-roU buoy at Station 13, operated from "Moray Eh-th" (DC, JE, P. Collins, C. Clayson). A second pitch-roll buoy was loaned to the Deutsches Hydrographisches Institut. Westhighouse Research Laboratories. 1968: Six-instrament array at Station 4, pressm-e transducers at Stations 1 and 3 (R. Bower, H . Martin, TB). 1969: Sis-instrument array at Station 4 (R. Bower, H . Martin, G. Bowman, TB). The wave instrumentation and telemetry were built by H . Martm, Ocean Applied Research Corp., San Diego. During the experhnent the telemetering Stations 1-5, 7-10 were serviced by the tugs " E h r " (1968) and "Eriechioh Voge" (1969), -with support from ah-sea rescue launches based in the Bundesmarine Station i n List and occasionally the life vessel "Hindenbm-g". The prelhninary editing and spectral analysis of the data was carried out by each group individually. A general system for assembling and processing the complete set of spectra * Liitials refer to authors.
26
Erganzungsheft zur Deutschen Hj^ckographisohen Zeitschrift. Reihe A (8), Nr. 12, 1973
was developed by WS. The analysis and interpretation of the data was then carried out chn-ing two workshops at the Woods Hole Oceanographio Institution from February to Apr, 1971 (WS, K H , JE, D H , PM, T B , E B , HC) and May-August, 1971 (WS, K H , E B , HC, DO, PM). Computations of the nonlinear energy transfer rates were made on the CDC 6600 and 7600 computers at the National Center of Atmospheric Research i n Boulder (WS, K H ) . Woods Hole Ooeanographie Institution provided assistance dmdng the preparation of a first di-aft of this paper (KH)*. Acknowledgments The work was supported by the Mhfisterimn f r Bdung und Wissenschaft, the Verteidigmigsmhiisterium, and the Verkehi'sministerimn (ERG); the Office of Naval Research, the National Science Poundation [NCAR], and the Woods Hole Oceanographio Institution [Doherty Eund] (USA); the Natmal Envh-onment Research Counc and the British Ministry of Defence ( U K ) ; and the NATO Science Commhtee. Ships were made available through the Verteidigung.sministerium, Bundesmarine ("0T2", "Holms", "Phi-", "Eriechich Voge", ah sea rescue boats); the U K Hydrographie Dept., M.O.D. (HMS "Enterprise" and "Bulldog"); the U K Natural Environment Research Council (charter for "Moray Pirth"); the Wasser- und Schiffahrtsdhektion Hambm-g ("Waher K r t e " ) ; the Deutsches Hydi-ographisohes Instfiut ("Gauss"); and the Deutsche Gesehschaft zm- Rettung Schfbrchiger ("Hindenbm-g"). We are grateful to these agencies and the individuals mentioned i n the previous section, and numerous others not listed explicitly whose generous assistance and support made this project jDossible. * Tliis paper also appears in the Collected Reprints of the Woods Hole Oceanogr; Institution as contribution no. 2911.
2. Wave Growth 2 .1. Generation Cases The primary pm-pose of the wave generation measm-ements was to determine the som-ce hmotion S' i n the spectral energy balance (radiative transfer) equation* dF dF , + Vi =S' (2.1.1) dt 3.^^ for the two-dimensional wave spectrum F(f,9; x, t), where ii. = = ^ is the group velocity. "^î ^^f ^'^ Much of the discussion mil, i n fact, center on the contracted form of this equation obtamed by integrating over the propagation dhection 6, dE at + ^ dE dxi =S (2.1.2)
where E = E{f; x, t) and S are the one-dimensional frec[uenoy spectram and soiu-ce function, respectively, and = ^VjF AdfE. I n either eciuation the som-ce function can be determmed from wave data taken along the profile only i f the gradient of the wave spectrum transverse to the profile is small compared -with the derivative parallel to the profile. Tor an ideal geometry consisting of a profile perpendicular to an infinite straight shore, this will be the case i f the field is homogeneous with respect to the parallel-to-shore coordmate x^ = y. The dhection of the -wind and the -ivhid-field variations -with respect to the coordinate x-^ = x parallel to the profile or time t may be arbitrary. However, i n practice, the length of straight shoreline perpendicular to the profile was finite (cf. Figs. 1.1 and 1.2) and we felt reasonably confident that transverse gradients could be neglected i n Eqs. (2.1.1) or (2.1.2) only for offshore muds blowing within about + 30" of the profile dhection. Although no restrictions were necessary with respect to the fetch or time variabfiity of the wind field, the conditions on the wind dhections were then normally satisfied only for wind fields which were also fahty homogeneous with respect to X and t. S. A. K i t a i g o r o d s k i i [1962] has suggested that under these conditions the wave data for different wdnd speeds and fetches should be expressible i n terms of a shigie nondimensional fetch parameter x = gxju^, where u.^, = (XIQ^Y'- is the friction velocity and T is the momentum transfer across the ah-sea interface {Q^ = density of ah). We found K i t a i g o r o d s k i i ' s relation to be fahly well satisfied, and the som-ce functions computed for indi-^ddual cases agreed reasonably well with the mean som-ce fmiction inferred from the ^'-dependence of the wave spectra after averaging over all "ideal" cases of stationary, homogeneous wind fields. Many of our general conclusions have accordingly been derived fi-om the mean spectral distributions after scaling with respect to nondimensional fetch (Sections 2.3-2.6). However, a number of indi-vidual generation cases were also analyzed (Section 2.7) to indicate the variability of the data and the representativeness of the mean results, inclucling possible limitations of K i t a i g o r o d s k i i ' s sealing law. I t appeared that most of the observed variability of our data was associated -with the small scale gustiness of the wind field rather than systematic deviations from K i t a i g o r o d s k i i ' s law. Nevertheless, there is some indication that further parameters not included i n K i t a i g o r o d s k i i ' s analysis, such as boimdarylayer scales, must be taken into account i n comparing laboratory and field data. * We ignore refractive effects in tliis section, but will consider the full equation in Part 3.
28
Erganzmigsheft ziu- Deutschen Hydrographischen Zeitsclu'ift. Eeilie A (8"), Nr. 12, 1973
2 . 2 . Kitaigorodskii's Similarity Law K i t a i g o r o d s k i i has pointed out that i f the geometry of the wind field and the laws of wave generation are sufiieiently simple that the wave spectrum can be imicpiely specified by a single fetch parameter x, the local friction velocity u.. and the gravitational acceleration g, then by dimensional arguments the spectrum must be of the general form F(f,0) = g^r'F(f,0;x) (2.2.1)
where the parameters = fu^jg, x = (7.^/w| and the function are non-dimensional cpiantities. Clearly the general three-dimensional space-time dependence of the wave field w i l l reduce to a single fetch parameter x i f the wind field and boundarj^ conditions are stationary and homogeneous vdtli respect to the dhection y perj)endicular to x, e.g. for a constant offshore -ndnd blowing perpendicular to an iirfinite straight shore. However, i t is less obvious that the spectrum shoidcl depend i n addition only on the l o c a l fi'iction velocity (besides g). Althotigh i t appears reasonable to assume that the friction velocity can adec(uately characterise the local interaction between the atmosphere and the wave field - a.t least to first order - this is relevant only for the local rate of change of the spectrum. The siiectrum itself must be determined by integrating the radiative transfer ecjuation and thus represents a net response to the enthe upwind m n d field. To avoid this difficifitjr, one could assume that M,,, is constant, i.e. independent of fetch. Although this finds some observational support, i t is difficult to find a convuicmg physical reason why this should be the case a priori, uifiess i t is assumed that the wave field plays a negligible role i n the momentum transfer across the ah-sea interface - which is not supported by experiment. I t appears more consistent to regard instead as an internal variable of the problem which has to be determined as a function of fetch i n the same way as the spectrum. Adopting this viewpoint, K i t a i g o r o d s k i i ' s spectrum (2.2.1) can nevertheless be derived hj the foUoAvhig, slightly revised argiiment. As relevant external parameter we speoifj' the constant "wind speed at some height above the sea surface which is no longer significantly affected by the wave field - ideally, outside the planetary boundary layer. The wave spectrum and u.^, are then governed bj'' the couiiled radiative transfer and atmospheric boundary-layer ecjuations under appropriate u p m n d initial conditions. Assummg that the wave spectrum is zero at the upwmd shore line x = 0 and that at a sufficient distance from the coast the initial properties of the boundary lajêr at x = 0 are no longer important, i t follows that the spectrum and friction velocity can be fimctions of x, U and g only. Hence by the same dimensional argument as employed by K i t a i g o r o d s k i i , the spectrum must be of the general form (2.2.1) with / and x replaced by /' = fUJg and x' = gxjU^. Shnilaiiy, i'JUo} must be a universal fitnetion i.i{x') of the nondimensional fetch x'. However, both scaling laws are equivalent, as is seen by mdting J'(/, 0, x) =F[f'n, 0, x'j.r'^) = f (/', 0, x'), say. Clearly, the same argument would have applied i f had been replaced by some other characteristic bouudary-lajêr velocity, e.g. the velocity at 10 m height above the smface. Since U-^^ is easier to measirre than and was routinelj' determmed at most stations along the profile, we have sealed all om- data i n the following with respect to U^^g rather than tt,.,, denotmg the appropriate variables by .r = g.r/Z7JQ, f = jU-^Jg, etc. Por comparison with data from other authors, however, we have also inchxded axes for the corresponding variables scaled with respect to u... using the relation M | = c^g [7jg Avith a constant drag coefiieient = 1.0 X 10-3. 2.3. Wind Sti'ess and Tm-bidcuce Measm-ements Dm-ing the main 1969 experiment, chag coefiicients were determmed at Station 8 from turbulent momentixm fiux measurements using both hot-whe and otxp-and-vane anemometers (cf. 1. 3. A more detailed description of the apparatus is given i n K . E n k e [1973]). Eor neutral or stable conditions the average value of CK , was found to lie f a h l y close to the value 1.0 X 10^^
29
10"
10' _l
10-
10'
10
10'
O UPPER ANEMOMETERl
STABLE
L O W E R A N E M O M E T E R ! STRATIFICATION
6.H
D UPPER ANEMOMETERl UNSTABLE LOWER ANEMOMETERl STRATIFICATION
L I U (1971) W E I L E R e BURLING (1967)
2.H LARGE KRAUS (1968) FETCH
(TABLED
10
to
10
10
10^
Fig. 2.1. Drag coetficients c, versus non-dimensional fetch x= gxl U'. Large-fetch measurements are listed in Table 1
Fig. 2.2. Reynolds stress sprectrum [k = 2ii U L/,,
= anemometer height, U
U,)
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973
31
applied i n the previous section. Under these conditions no systematic dependence of the drag coefficient on x was found, cf. Mg. 2.1. Slightly larger di-ag coefficients were found for unstable ah-sea temperature differences, "with indication of a decreasmg CJQ trend vdth increasing x. Since most of our wave data were obtained under stable conditions, we have ignored the temperature dependence i n convertmg from l /^p to u.^. This procedme is consistent "with K i t a i g o r o d s k i i ' s scaling law, which also disregards temperatm'e elfeots. Figure 2.1 and Table 1 indicate considerable v a r i a b i t j ' i n the drag coefiicients measured by difi'erent groups. This causes difficulties i n applying K i t a i g o r o d s k i i ' s scaling law, which can be expected to be valid only i f C^Q is a miic[ue function of the non-dimensional fetch X. I n the literature, wave parameters are tisually presented in nondimensional form with respect to u,^. To compare wave data from other sources with orrr data i n Figs. 2.6, 2.7 (Section 2.4) we have simply plotted all parameters with respect to our axes - although the chag coefficient Cjg = 10"^ used to convert from U to may differ from that used i n the original source. Slightly different plots woid have resulted i f all data had been nondimensionalized vdth respect to Z/JQ. (In practice, however, variations of CJQ by a factor of 2 or 3 are not very noticeable i n the log-log plots of Figs. 2.6, 2.7.)
Table 1 Large Fetch Values o c
Author Smith [1967] Smith [1970] Hasse [1968] Brooks et al. [1970] DeLeonibus [1971] Pond et al. [1971] JONSWAP
Cio
10^ m/s 3-13 6-15 3-11 3-13 3-15 4-7 3-11
Stratification
Method correlation
0.9 0.24 1.35 0.34 1.41 0.26 1.20 0.48 1.21 0.24 1.300.18 0.8 0 . 4 1.2 0 . 4 1.3 0 . 6 1.520.26 1.0 0 . 4 1.2 0 . 4
stable
correlation correlation
neutral stable neutral unstable stable unstable
profile correlation correlation correlation
The mean spectrum of the shear stress and its r.m.s. variation over the set of measm-ements are presented i n Fig. 2.2. Figures 2.3 and 2.4 show the corresponding spectra for the horizontal and vertical velocities. The distributions agree Cjuite well m t h measurements by other workers. Cross spectra between atmospheric tm'bulence fiuetuations and the sm'face displacements, which were measmed at the same position, were also analyzed, but no significant correlations were detected. The ratio anemometer height to the wavelength of the principal wave component was typically of the order 0.2. J. E l l i o t t [1972] also found no correlation for comparable ratios*. I t appears that turbulence measurements considerably closer to the sm'face are needed to study wave-induced velocity fiuetuations.
* However, both E l l i o t t and we observed correlations for swell comîonents whose phase velocities exceeded the wind speed - as have a niunber of other authors. The phases were in accordance with potential theory.
32
Erganzungsheft zm- Deutschen Hydrogi-aphischen Zeitschi-ift. Eeihe A (8), Nr. 12, 1973
2.4. Petcli Dependence of One-Dimensional Spectra The fetch dependence of the one-drmensional frequency spectra were investigated by parametrizing them with a least-square fitted analytic function. Various functional forms were tried. A uniform good fit to nearly all of the spectra observed during "ideal" generation conditions was finally attained by the function (/ - i J 2cr^/;
E (/) = ag' {2n)-*
exp ( - ; { ^ Y ) y ' " ^

(Ta
(2.4.1)
for
contahnng five free parameters a, and a^,. Here represents the frequency at the maximum of the spectrum and the parameter a corresponds to the usual P h i l l i p s constant (the spectrum approaches the usual f-'^ power law for large //,). The remaining three parameters define the shape of the spectrum: y is the ratio of the maximal spectral energy to the maximum of the corresponding P i e r s o n M o s k o w i t z [1964] spectrmn* 5 / (24.2)
-with the same values of a and f^; and define the left and right sided widths, respectively, of the spectral peak. The functional form (2.4.1) was obtained by midtiplying a P i e r s o n - M o s k o w i t z spectrum with the "peak enhancement" factor
The pm-pose of the parametrization was not to contribtite a fm-ther emphical spectrum, but rather to reduce the 100 frequeney points of the measured spectra to a manageable, nonredmidant subset of parameters whose dependence on fetch, windspeed and other factors could be systematically investigated. (The parametrization also had advantages i n differentiating the spectra to determine the source functions, cf. Sections 2.6, 2.7). Pormulae for fetch-limited spectra proposed by other authors, while perhaps adequate to describe certain properties of mean spectra, were fomid to contain too few parameters to characterize the variations encountered i n the individual measm-ed spectra. Pig. 2.5 shows a typical series of spectra measm-ed under "ideal" generation conditions, together -ndth theh- parametrical approximations. Plots of the five spectral parameters for the 121 spectra measured under "ideal" generation conditions agahist nondimensional fetch x = gx\'U\g are shown i n Pigures 2.6, 2.7 and 2.8. Pigures 2.6 and 2.7 also include scale parameters /, and a from other experiments. The distribution exhibits the smahest scatter. I n part this may be due to the fact that , is the most accm-ately defined spectral parameter. However, the smaller variability of the distribution as compared m t h the a data or the still more scattered shape parameters can also be understood from the origin of the scatter, which we attribute i n Section 2.9 to the gustiness of the wind. * The exponent of thePierson-Mosliowitz spectrum is normally -written as 0.74 (//)"'' with reference to the frequency = gl-KTJ.^^ for winch the phase speed of the waves is equal to the -wind speed. The two expressions are identical, but in our case - or equivalently - is treated as an adjustable parameter.
33
0.7-,
Hz
Fig. 2.5. Evolution of wave spectrum with fetch for offshore winds (11^ -12^, Sept. 15,1968). Numbers refer to stations (of. Rg. 1.2.). The best-fit analytical spectra (2.4.1) are also shown. The Inset illustrates the definition of the five free parameters in form (2.4.1)
34
10^ l_
10"
10=
10'' I
10' I
x=xg/U,o
5H , 2
205
1 Burling(1959) - Pierson(1960) " KitaigorodskiitStrekolov(1962 PiersoncMoskowitz(l9e4) Hidye Plate (1966) " Mitsuyasu (1968) Volkov(1968) Kononkova(1970) Liu (1970) Pl^Pierson Moskowitz (1964) + This Experiment
02
01
PM
Fig. 2.6. Peak frequency v. fetch scaled according to Kitaigorodskii. Data at small fetches are obtained from wind-wave tanks. (Capillary-wave data was excluded where possible.)
10'
10'
2
.1
a
05
ABurling(1959) i. Pierson{1960) V Hicks (1960) 0 Kinsman (1960) Longuet Higginsct a((l963) X Sutherland (1967) o ( |itsuyasu(1968) Liu (1970) o Toba (1971) PM Pierson Moskowitz (196A) ^ S c l i u l e e t a l (1971) Pierson & Stacy (1971) .005 I + Ttîs Experiment 02 h 10" 10^ 10" 10=
1tf'
10
Fig, 2.7. Phillips constant v. fetch scaled according to Kitaigorodskii. Small-fetch data are obtained from wind-wave tanks. (Capillary-wave data was excluded where possible.) Measurements by Sutherland (1967) and Toba (1971) were taken from W.J. P i e r s o n and R A S t a c y [19731
PEAK SHAPE PARAMETERS

/!
10'
I
10=
0.3
10^
10=
0.2
0 1
0.0
0.2
0.1
H
+ + + ++ +
0.0 10^ 10-
IQ
10'
.37
Oiu- a parameters i n Fig. 2.7 are seen to lie a little higher than the data h-om some of the other groups. This may be due partly to the function-fitting technicûes used to evaluate a. H . M i t s u y a s u [1969] for example, mentions three methods, of which the one used for the a values shown i n Fig. 2.7 yielded systematically smaller estimates by 10 to 20% than the other two. Om' technic[ue of simultaneously optimizing five spectral parameters tended to yield a values about 20 % higher than, say, a straight-line fit to the function E (/) f for all frec[uencies / greater than f^, which corresponds more closely to the standard method. A significant featiu'e of Fig. 2.7 is that a is clearly n o t a universal constant, as requhed by P h i l l i p s ' original dimensional argument. Om' analj^sis of the source functions presented later i n Sections 2.6-2.9 indicates that the energy balance i n the region of the sj)ectrum between and about 3 , is indeed not governed by white-capping, as envisaged by P h i l l i p s , but primarily by the energy input from the atmosphere and the nonlinear energy transfer by wave-wave interactions towards lower and higher frequencies*. The straight line /, = 3.5 x-"-'' (2.4.3)
in Pig, 2.6 represents the best fit to the data (H. M i t s u y a s u [1969] finds the same exponent); the line a = 0.076 x--^^ (2.4.4)
in Fig. 2.7 is based on dynamical scaling considerations (Section 2.8). Although i t jdelds a reasonable overall fit for all field and wave-tank data, an exjionent of 0.4 is a better approximation to the average slope of om' a distribution. The peak shape parameters y, and (Fig. 2.8) are more highly scattered than the scale parameters. Since they characterize the properties of a rather narrow spectral peak, some variability is to be expected from the statistical indeterminacy associated "with the spectral measm'ement. However, this accounts for an r.m.s. relative variability of only about 20%, which is considerably smaller than observed. Attempts were made to reduce the variability by subdividing the data with respect to additional parameters such as the ah-sea temperatm'e difference, currents, and the dhection of the wind relative to the profile. No reduction of the scatter was achieved. However, local non-stationaries (defined by the time derivative estimated from three neighboring spectra) above a certain thr'cshold were found to account for some of the variabfiity, and Figs. 2.6, 2.7 and 2.8 accordingly contain only generation cases for which the non-stationaritj' was below a critical value, determined emph'ieally by the inecpiality ^ < where represents the group velocity of the waves at the spectral peak. ot dx I t ap23ears that the residual scatter of om' data i n Figs. 2.6, 2.7 and 2.8 is geophysically real and cannot be accounted for i n a simple manner by additional parameters characterizing the mean generation conditions. The analysis of indi-vidual generation cases discussed i n Section 2.7 shows that a variabfiity of the same order occurs vdthin each generation ease, suggesting that i t may be associated with small scale inhomogeneities of the TOnd field (Section 2.9). A clear trend of the shape parameters with fetch is not discernible within the scatter of the data, although there is some indication of y peaking to a maximum around x = 10^. Accordingly, we shall assume as first approximation that the spectral shape is self-preserving and consider i n much of the foUowing a mean spectrum defined by the parameters** y = 3.3, cr = 0.07, = 0.09. (2.4.5)
* I t could he sm'mised that the fetch dependence of om- a data is induced by the fetch dependence of o\a averaging band, defined with respect to e.g. due to an exponent in the spectral power law which diifers slightly from 5. This is not the case. Essentially the same results were obtained using fixed freciuency bands. ** A more detailed dynamical analysis suggests that this is not strictly tenable, cf. Seotion2.9.
38
Erganzungsheft zur Deutschen Hych'ographisehen Zeitschrift. Reihe A (8"), Nr. 12, 1973
Fig. 2.9 shows a comparison of the normalized shape functions (p(p\={2nta-'g-'f'E(f) (2.4.6)
for this spectrum, the P i e r s o n - M o s l i o w i t z spectrum and a schematic empirical spectrum proposed by S. A. K i t a i g o r o d s k i i [1962]. The form corresponding to (2.4.5) is cjualitatively very similar to the shape functions observed by T. P. B a r n e t t and A . J . S u t h e r l a n d [1968] and H . M i t s u y a s u [1968a, 1969]. To conclude om' summary of emphical fetch relationships. Fig. 2.10 sliows a jjlot of the nondimensional energy i = S X c/ 17^^,^, where S = 1 E (/) d/. The data is well represented by the linear relationship # = 1 . 6 x lO-'-T. (2.4.7) For a seh-simikr speetmm, S ~ so that the exponents Hj-, for a and S should satisfj^ the relation 4?ij.-H^ + ? j ^ = 0 . The values given i n equations (2.4.3), (2.4.4) and (2.4.7) yield 4(-0.33)-(-0.22) + ! = - 0 . 1 (2.4.9) and w ^ i n the power laws (2.4.8)
which lies within the error of about + 0.2 estimated from the error bands of the individual exponents. 2.5. The Dhectional Disti'ibution The majority of investigated generation cases ocom-red dm-ing the latter part of July 1969, when the dhectional data of all but one of the pitch-roll buoys was questionable, or dming the period August 1-August 15, 1969, when only the innermost section of the profile (without pitch-roll buoj^s) was i n operation. Unfortunately, the directional data from the single pitch-roll buoy and the linear array at Station 4 did not encompass a sufficiently wide range of dimensionless fetch values x to permit a systematic investigation of the fetch dependence of dhectional parameters. Thus we have assumed - i n analogy with the similar result fomid for the hequency spectra - that the scaled dh-ectional distributions s{6, fifj,,) are independent of fetch. Averaging over all ideal generation cases, the linear array yields a mean beam -width for frec[uencies near the spectral peak of 0^ 19 + 5". This corresponds to an ecûivalent pitch-roll beam width of 6^ x 27" + 8" (/ x 0.4 Hz) which is reasonably close to the value 31" for a cos^ 9 distribution. A t frequencies appreciably higher than the peak frec[uency, the linear an-ay was normally too strongly aliased to be interpreted reliably. The pitch-roll buoy at Station 10 also jdelded beam widths 9^ x 30" for f = f^. Examples of the dependence of beam width on frecptencies greater than is shown i n Pig. 2.11. Data from both the fimctioning and questionable pitch-roll instruments are included in the fig-ure. The malfunctioning instruments were characterized by random variations i n the mean dhection from one recording to another. Rather sm'prisingly, however, the mean scûare angular spread for all instruments were found to be reasonably consistent. A possible explanation is that the compass was sticking and the orientation of the buoj^ remained relative^ constant dm'ing any given recording. This interpretation finds some support through the fact that for the same wave conditions the frequency dependence of the mean angle 0, was generally similar for all pitch-roll instruments except for random shifts of the zero reference angle. We infer from Fig. 2.11 that the dhectional distribution maj' be modeled by cos^ 9 (0s K, 30") near the peak of the spectrum - i n accordance with the linear arraj' data changhig gradually to an approximately uniform half-plane radiation ( 0 ^ 4 3 " ) at high frequencies, > 2 to 3 .
30
60 a 's
30
i STA. 6 JULY 2 6 2 S T A . 6 JULY 2 7 3 STA, 7 J U L Y 2 7
3 0 22'
0 0,30 ' 0 0,30 '
Fig. 2.11. Typical frec[uency dependence of tîe root-mean-sc[uai'e directional sju'ead 0s measiu'ed for fetch limited spectra Since most of the wave energy is concentrated near the peak we have assumed for shnplicitjf a frequency independent eos^ 0 spreading function i n computing net som-ce functions, nonlinear transfer rates, etc. I t wâs found that the prhieipal conclusions regarding the energy balance were i n fact rather insensitive to the precise form assumed for the sjireading fimction. 2.6. The Mean Source Function The emphical average dependences (2.4.3), (2.4.4) of the spectral parameters and on fetch x = gxlUlg, together -with the (constant) average shape parameters (2.4.5), define a mean one-dimensional spectrum at each x, from which a mean one-dimensional som-ce fmiction can be computed through ecpiation (2.1.2), (2.6.1) Fig. 2.12 shows the mean JONSWAP spectrum and the associated mean som-ce function S (a cos^ 0 spreading factor was assumed i n evaluating v). The plus-minus distribution of the som-ce function is associated with the transition of the spectral peak towards lower frec[uencies, the positive lobe corresponding to the rapidly growing waves on the forward face of the spectrum and the negative lobe to the attenuation of the waves on the rear slope of the peak as they approach equilibrium (the overshoot effect - cf. also Fig. 2.5). Several authors have pointed out that the form of the som-ce function, i n particular the overshoot effect, is difficult to explain i n terms of a linear (or any frequency decoupled) theory, and have suggested that cross-spectral energy transfer due to resonant wave-wave interactions, white-capping or some other form of nonlinear couplhig may be significant (T. P. B a r n e t t and A . J . S u t h e r l a n d [1968], B . V , K o r v i n - K r o u k o v s k y [1967], H . M i t s u y a s u [1969], A . J . S u t h e r l a n d [1968]). Our computations of the nonlinear energy transfer due to resonant wave-wave interactions, also shown i n Pig. 2.12, indicate that this process can indeed account for the prmeipal features of the observed som-ce function, including the plus-minus signature, the major part of the wave growth on the forward face of the peak, and the overshoot phenomenon. We attribute the difference between the observed som-ce function S and the nonlinear energy transfer *S, to the positive energy input from the atmosphere, which is discussed fm-ther i n the context of the overall energy balance of the spectrum i n Section 2.9.
40
Erganzimgsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12, 1973
0.4 !-
0.0
0.1
0,2
0.3
0.4
0,5
0,6
0.7
0,8
0,9
1.0
Eig. 2.12. Mean JONSWAP spectrum, source function S computed from power laws (2.4.3), (2.4.4) using (2.7.1), and computed nonlinear energy transfer Snj
The spectra and source functions shown i n Fig. 2.12 correspond to particular values of the fetch x and windspeed U^^. I n general the ratio of the soiu'ce functions S and may be expected to vary with the nondimensional fetch x. However, under certain scaling conditions which we return to i n Section 2.9 - the ratio is approximately independent of x. From the Boltzmann-integral expression for the nonlinear energy transfer (K. H a s s e l m a n n [1962, 1963 a]) - namely
St X 8 {coi + (O2 - (O3 - 0)4) dfej dfeg where = hj^ + k^, Nj = N{kj) = {kj)lcOj, C0j=]/gkj, F{k) is the two dimensional wave spectrum with respect to wavenumber, and the transfer coefficient T is a ciuadratic homogeneous function of the four wavenumbers fej, . . . k^ - it follows that for a family of self-similar spectra of the form E'{f, 9) = agîZiz)-^ f-^ ^' dUm' ^) nonlinear transfer scales w i t h respect to a and according to S, = Ulg-'a'fZ'ilyUIL) (2.6.2)
where !/'(//) is a dimensionless function depending on the shape of the speetmm. To compare and 8 it is useful to consider integral momentum transfer properties of the functions, which are relatively insensitive to the shape of the spectriun (and w i l l be related later to the momentum transfer across the ah-sea interface). Since wave-wave interactions are conservative, the net momentum transfer Q g df is zero, {k^ represents the average
value of k^ integrated over the normalized dhectional distribution of /S,,,.) However, the integral over any one of the three lobes of the distribution i n Fig. 2.12 is non zero, and is of the general form, say for the t h h d lobe (i.e. the positive lobe at high frequencies),
41
/c27r/
(2.6.3)
where c is a dimensionless constant dependent only on the shape of the spectrum and represents the lower l i m i t of the lobe (cf. Pig. 2.12). On the other hand, according to (2.1.1) the net momentum transfer Q^g f - Sdf is equal to the momentum advected away by the wave field, o = Q-y9 J Qw9 (2.6.4)
2nf
dx
dx
The factor 3/8 follows from the assumption of a cos^ 6 spreadhig factor. Emphically, S' ~ x^^-^ from (2.4.7), /, ~ from (2.4.3), so that the ratio Thf/r^j remains constant i f a ^ .;-o.22o.oi as given by (2.4.4), i n accordance with the overaU cUstribution of a. i n Tig. 2.7. We note that these exponents correspond not only to a constant ratio T h f / t a d , but also to fetch independent equivalent friction coefiicients Cjjf = T^Je^ Ulg, c^^ = T^JQ^ U^^. The ratio Tt^jT^^ is rather insensitive to the angular distribution. Tor example, a cos* 6 rather than a cos^ 9 spreading factor increases by about 10%, whereas the computations of Si yield about 15 % higher values for TÎ . The uneertahities associated m t h the angifiar distribution are generally smaller than the variabfiity introduced through the cubic dependence of (eq. (2.6.2)) on the rather scattered parameter a. However, a eos^ 9 distribution near the peak is consistent w i t h computations of the two-dimensional source fimction Sî, which has approximately this angular distribution i n the low-frec|uency positive lobe (cf. Fig. 2.13). A narrower angular beam is fomid to become broadened by the noifiinear transfer near the peak. Fig. 2.13 also indicates that the nonlinear transfer tends to broaden the angular distribution at higher frecjuenoies more than at low frecjuencies, i n accordance wdth Fig. 2.11. As pointed out i n the previous section, the power laws (2.4.3), (2.4.4) and (2.4.7) are almost but not strictly compatible with a seh-simfiar spectrum, since they violate (although by less than the error limits) the condition (2.4.8)*. We retm-n to the question of self-similarity and scaling i n comiection with the overall momentum and energy balance of the spectrum
Fig. 2.13. Two-dimensional nonlinear som-ce function S (, 9) computed for the mean JONSAVAP spectrum * Strict self-similarity together with a constant ratio
X. Thf/lâd
requu-es
X'
42
Erganzvingsheft zur Deutschen Hydi'ographisohen Zeitschrift. Beihe A (8"), Nr. 12, 1973
i n Section 2.9. We simply remarli here that the spectral scales vary i n such a manner that to first order both the som-ce fimction ratio S:Si and the nonlinear momentum transfer T,,f to high frequencies remam constant over a five-clecade range of fetches. This is apparently not coincidental but a consecpience of the nonlinear interactions. 2.7. Analysis of Individual Generation Cases To establish the representativeness of the somce function derived from the mean fetch dependence of the set of all scaled spectra, five generation cases were investigated independently. The examples were chosen arbitrarily from the various periods of "ideal" generation conditions. I n each case the spectral parameters (/,, a, y, a^, a^,) = (a^, a^, a^, a^, a^) were evaluated for each spectrum and plotted as fmictions of fetch x and time t. From the spatial and time derivatives of the parameters obtained graphically from these plots the somce fmiction was then determmed using the relation , , . M + ,9^=9^f9^ + ,9^) dt 9.r dUjXdt dxj (2.7.1)
for the parametrised spectrum. This technique was regarded as more accurate than straightforward estimates of the spatial and time derivatives from finite differences of the measured spectra. On account of the peaked form of the speetmm, the spatial and time resolution of the s]peotral field was oifiy marginally adecpxate to trace the rapid growth of a given frec|uency component dm-ing its brief period on the forward face of the spectrum. Thus i f the finite difference method is applied i n the usual manner to estimate dF ^^F^ 9i ~ A i ' dl[^ ^F
9a; ~ Aa;
at a fixed freciuency, the soiu-ce function appears strongly smoothed with respect to frequency. This difficulty is avoided by transformhig to a new representation i n terms of parameters . . . a which vary smoothly with respect to space and time. A comparison of the source fmictions computed by both techniques is given i n Fig. 2.19 (generation case 4). A listing of the relevant parameters of all cases studied (includhig three other JONSWAP spectra and the Pierson-Moskowitz speetmm) is given i n Table 2. The weather situation for all generation cases was very similar and is illustrated by the two examples i n Figs. 2.14 and 2.15. More detafied histories of the -ivinds measm-ed along the profile, together with the times and positions of the wave measm-ements, are shown for the individual generation cases i n the appendis. Fig. 2A.1. The fetch and time beha-viom- of the spectral parameters are presented in Figs. 2 A . 2 - 2 A . i l of the appendix. I n all cases, the x dependence of the scale parameters could be adequately represented by the power laws / ~ x-"-^, a ~ ft;-"-*. The shape parameters were too strongly scattered to define derivatives, so that these were taken to be zero, as in the average case. Figm-es 2.16-2.20 show the source functions, 8 evaluated for each case at the positions indicated i n Figs. 2 A . l - 2 A . i l using equation (2.7.1), and the computed noifiuiear transfer rate SîAlthough both the details and relative magnitudes of S and show considerable variabfiity, the resifits tend to confii-m the average som-ce functions of Fig. 2.12. The variability appears consistent with the inaccm-aeies entaed by the experimental differentiations and the sensitivity of the cubic wave-wave interaction integrals with respect to the wave spectrum. We note that the scatter of the spectral shape parameters is characterized by rather small spatial and time scales and is comparable to the scatter seen i n the plots for the complete ensemble of generation eases (Fig. 2.8) - suggesting that i n both cases the variability is caused by the gustiness of the wind.
Fig. 2.16. Spectrum, empirical source function and computed nonlinear energy transfer for generation case t Fig. 2.t7. Same as Fig. 2.16 for generation case 2 Fig. 2.18. Same as Fig. 2.16 for generation case 3 Fig. 2.19, Same as Fig. 2.16 for generation case 4. Ttie inset shows the (less accurate) estimate of the source function obtained froi direct differences of the measured spectra. The spectrum represents the average of the three spectra used in the finite differences; the nonlinear transfer rate was computed for this mean spectrum Fig. 2,20, Same as Fig, 2,16 for generation case 5
46
1
, .
I
0.0
I
0.4
I
0.8
1
.
(a)
Hz
(b)
SHARP SPECTRUM
Fig. 2.21. Nonlinear energy transfer computed for four spectra wittl increasingly narrower peaks
Table 2 Generation Cases
Fetch
Time
Case
Date
O 1-H a b
. (
O b
1 o
o. E
SI H ro Iro B SI 1^ ro 1 ro
ro|fo ^= E| ror"
ro
O u a
i * ^ > II n O c o
1 2 3 4 5 Mean JONSWAP Sharp JONSWAP Very sharp JONSWAP PiersonMoskowitz (1) aP,
(2) =
12 15">
15 Sep. 1968 26 Jul. 1969 1 Aug. 1969
km 37 11 5.25 8 5.25
m/s 7.0 8.8 9.0 10.7 9.2
9.1 15.6 15.5 20.4 20.5
Hz 0.263 3.06 0.371 0.421 0.349 3.67 3.42 3.86
85 104 72 57 60 70 50 68
83 104 92 77 100 90 70 123
7.4 1.39
0.188
10.6
0.187
0.92
0.36 0.046 -0.24 0.045
0.08 0.15 0.09 0.22 0.18
0.13 0.20 0.14 0.29 0.25 0.18
0.333 15.4 18.3 18.0 18.5
0.325 0 + 0.05 0.421 0.411 0.428 0.16 -0.27 0.12 0
0.636 0.386 0.685 0.381 0.608 0.390
0.18 0.052 0 + 0.02 0.072 0.16 0.072 0
1930 1530
1 Aug. 1969 10 Aug. 1969
0.416 3.24 3.3 4.2 7.0
0.050(3) 0.13 0.12
8.1
> 20- 0.14 50
0.09
0.09
as inferred from power laws (2.4.3), (2.4.4) for given x = gxjXJ+ CM-' T " ' 7 T = c'"/Ci = 10' for c^, = lO'K
(3) From equation (2.4.7)"
48
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12, 1973
2.8. Origin of the Spectral Peak I t appears from the preceeding two sections that nonlinear wave-wave interactions can accomit for the observed shift of the spectral peak towards lower frequencies at about the rate observed. However, the evolution of the peak and its persistence throughout the development of the wave field still remain to be explained. To investigate this ciuestion, nonlinear transfer rates were computed for a number of spectral shapes with broader or narrower distributions than the mean JONSWAP spectrum. Pig. 2.21 shows the noifiinear transfer rates computed for a seciuence of spectra with successively sharper peaks. A cos^ 0 spreading factor was assumed i n all cases. Panel a gives an example of a fahly white spectrum wdth a low frequency cut-off. I n all such cases i t is found that the nonlinear energy transfer tends to enhance the energy level towards the low frec[uenoy end of the distribution, caushig a peak to develop. The evolution of the peak continues up to and beyond the already rather pronounced form shoivn i n the second panel (a P i e r s o n - M o s k o w i t z spectrum). This is evidenced by the position of the positive lobe of the som-ce function dhectly beneath the spectral peak. However, as the peak conthiues to shaiqieii, the positive lobe of the noifiinear energy transfer gradually moves towards the forward face of the spectrum, until the spectrum reaches a state (approximately given by the mean JONSWAP spectrum) where the peak no longer grows but shffts towards lower frec|uenoies Avithout change of shape (panel o). Por a still more pronounced spike, the nonlinear som'ce function finally develops two positive side lobes on either side of a strong negative lobe dheetlj^ beneath the peak, and the distribution tends to fiatten again (panel d). Hence i t appears that the evolution of a rather sharp spectral peak of approximately the form observed can be explahied as a seff-stabfiizing property of the noifiinear interactions. The superposition of an additional atmospheric energy input with a smooth freciuency distribution, as discussed i n the foUo^vuig section, does not appreciably affect the form of this self-adjusting spectral distribution. We have not been able to find a simple physical explanation for the particular forms of the nonlinear source fmictions shown i n Fig. 2.21. Generally, simifitaiieous conservation of two independent scalar ciuantities, i n this case energy and action, requhes a plus-miiius-plus distribution of the one-dimensional noifiinear energy transfer (K. H a s s e l m a n n [1963a] this is well kno-wii also for the case of two-dimensional turbulence, where the noifiinear interactions conserve both energy and mean square vorticity). However, the jiosition of the first jiositive lobe dfi'ectly beneath the spectral maximum for moderately peaked distributions is rather miexpeeted, as i t is known fr'om general entropy theorems (in the present context of. K . H a s s e l m a n n [1963a] and [1966]) that wave-wave interactions change the spectrum hreversibly i n the direction of a uniform energy distribution i n wavenumber space (correspondhig to an one-dimensional frequeney spectrum). Intuitively, we should therefore expect a negative rather than positive lobe of the source function at the position of the peak. However, this is the ease only for extremely peaked spectra. Numerical investigations of the transfer integral indicate that the form of the transfer fmiction near the peak is dominated by interactions i n which all components of the interacthig quadruplets lie i n the vicimty of the spectral peak, and that the high-frec[uency range of the spectrum is of only minor significance for the evolution and stabfiity of the peak (cf. also K . H a s s e l m a n n [1963b]). A more detafied explanation of the behaviour of the computed iioiilhiear transfer distributions requh'es investigation of the rather complex algebraic sti-uotm-e of the coupling coefficients and resonant interaction smfaces. A complete compilation of the nonlinear transfer computations carried out for this study can be fomid i n W. Sell and K . H a s s e l m a n n [1972]. Several improvements over the numerical method used i n earlier computations (K. H a s s e l m a n n [1963b]) were introduced, including a symmetrical treatment of the wave "coUisioii" processes which yielded accurate overall conservation of energy, momentum and action, and furthermore provided consistency checks through three independent estimates of the transfer rates. Typical computation times on a GDC 6600 (NCAR) were 30 minutes per spectrum. The results were checked against
4-9
independent compntations nsing a different numerieal teclinique by D. C a r t w r i g h t (unpublished); the overall agreement of the smoothed distributions lay withm a few percent cf W S e l l and K . H a s s e l m a n n [1972]). The fine scale variabfiity of the nonlinear computations presented i n Sections 2.6-2.8 is associated with the fiifite mesh size used for the integration. Reduction of the mesh size by a factor of two removes most of the variabifity, but would requh-e an eight fold increase i n computmg time. 2.9. The Energy and Momentum Balance of the Spech'um Having gained some insight into the role of nonlinear interactions i n controhing the shape of the spectrum and fis evolution from short to long waves, we may attempt to construct a picture of the overall energy balance of the spectram. Unavoidably, we shall be handicapped herem by incomplete knowledge of the remaining processes governing the spectral energy balance - i n particular the input by the m n d and the dissipative mechanisms. Tims to some extent om- conclusions Avill remam ambiguous, and inequaUty relationships Aâll normally take the place of equations. However, i n certain cases two-sided inequahties can be narrowed down to yield quantitative predictions. n The net som-ee function S i n the energy balance equation may be divided generally into three terms,
representing, respectively, the energy input from the atmosphere, the nonlinear crossspectral transfer 8,^ due to conservative wave-wave interactions, and a dissipation term b^,. We include i n S,, not oifiy pm-e energy losses but also any cross-spectral energy transler which may accompany nonlinear dissipative processes such as whfie-cappmg, lor example i n ooniunction with the attenuation of long waves by damped short waves e 0. M P h i l l i p s [19631 M S L o n g u e t - H i g g i n s and R. W. S t e w a r t [1964], K . H a s s e l m a n n [1971]^ Prom the measm-ed net source function 8 and the computed nonlinear transfer we may infer the value of the sum 8, + S,, = S - S but can say nothing about the partition of the sum mthout fm-ther irformation about 8;, or 8,,. Assuming for the moment that the dissipation is negligible i n the mam part of the spectram, the som-ce fmictions 8 and as computed for the mean JONSWAP spectrum i n Pig. 2.12 imply a net energy balance of the form indicated schematically i n Pig. 2.22. The energy input has approximately the same distribution as the spectrum (in accordance vdth the linear dependence on the spectrum predicted by most wave generation theories) and is balanced by the negative noiiliiiear transfer away from the main part of the spectrum to lower and higher frequencies. A t low frequencies the posfiive noifiinear energy transfer is removed by advection, whereas at high frequencies some form of dissipation needs to be invoked to balance the positive contributions from both the noifiinear transfer and the atmospheric input. I t is of interest to compute the minunal momentum transfer T'" fi-om the atmosphere to the wave field needed to mafiitain this ''nfinhnum-dissipation" spectral balance, im-st, the atmosphere must supply the momentum
advected away by the wave field. I n Section 2.6 this was shown to be given by 3/8 e^g d#/da(eq. (2.6.4)), or substituting the emphical linear fetch dependence (2.4.7) for the energy ^ ,
where
c,, =
1.6 X 1 0 " ' = 0.05 X 1 0 ^ .
50
Erganzungsheft zm- Deutschen Hydi-ographischen Zeitschi-ift. Reihe A (8), Nr. 12, 1973
Fig. 2.22. Schematic energy balance for the case of negligible dissipation in the main part of the spectrum Secondly, the momentum T,,f transferred to high frequencies i n the positive lobe of the nonlinear source function, -\vhich is passed on to currents through dissipation, must be replaced by an equal atmospheric input -which enters the wave field i n the central region of the spectrum. This contribution was shown i n Section 2.6 to be of the form (eq. (2.6.3)) Ti,f = c^,^ a? , which reduces i n the case of the emphical power laws (2.4.3) and (2.4.4) to " i^hf = where % = cew/ea (0.076)^ ( 3 . 5 ) ( 2 . 9 . 3 ) Computations for the mean JONSAVAP spectrum yield Cj,f = 0.13 x l O - l Summing (2.9.1) and (2.9.2) we obtain for the minimal momentum transfer from the atmosphere to the wave field wfih C" = Cad + CM X 0.2 X 10-^ (2.9.4) ea U% (2.9.2)
Comparing this value m t h the drag coefiieient c^gXlQ-^ for the total momentum transfer T across the ah-sea interface, we conclude that at least 20 % of the total sea surface di-ag is induced by the wave field. We note that this represents a conservative lower l i m i t even i f dissipation i n the central region of the spectrum is discounted. I t assumes that the atmospheric input is identically zero for frequencies i n the high-frequency positive lobe of the nonlinear som-ce function, and conversely that the dissipation is ideifiically zero outside the lobe. A continuous transition of both source fmictions and S^^ across the lobe boundary, which appears more probable (cf. Fig. 2.22), would increase the estimate of the net atmospheric input. Assuming an arbitrary dissipation, rather than the minimal dissipation compatible vnth S and 8i, the wave induced drag coefficient c may take any value between c"'" and the theoretical maximum Cjp. The calculations for the coefiicients c^^, c^f and c'" apply for all fetches for which the power laws (2.4.3), (2.4.4) and (2.4.7) remain valid and the spectrum retains the same shape.
51
The latter condition is not strictly satisfied (cf. Section 2.4) but is not critical for the estimates. I t enters only i n the comiiutations for c^f through the shape dependence of the coefaoient c i n eq. (2.9.3). Tor the six spectra listed i n Table 1, C|,f varies by only a few percent. However, the calculations do depend c r i t i c a l ^ on the power laws of the spectral scale parameters. I n view of the cubic dependence of /S', and -c^f on a, the systematie deviations of our a data (Tig. 2.7) from the mean line (2.4.4) drawn thi-ough the overall data cannot be ignored. I n Tig. 2.7, dotted lines have been drawn parallel to the mean a line to delineate the theoretical upper l i m i t c,,t = Cjo - c^^ = 0.95 C^Q, which is five times greater than mean value Cijf = 0.18 0 ^ , and a corresponding value c^j = 0.036 c^j, which is one fifth of the mean value. The JONSWAP K data spans both limits i n the range 10^ <x< 10*, corresponding to changes i n Cj^j by a total factor of 25!* On the other hand, since the energy data (Pig. 2.10) fall along a fahly well defined line ~ x i n this range, the adveetive drag coefficient remains approximately constant. This forces us to reexamine our first-order hypothesis of selfsimfiarity and a common scahng law for S and Si. The ratios Tjf :rf iT^f of the momentum fluxes obtained by integrating the nonlinear (momentum) source function over the low-frequency (positive), median-frequency (negative) and high-frequency (positive) lobes, respectively, are generally of order 1: ( - 4 ) : 3. These values are insensitive to the details of the spectral shape. Por the mean JONSWAP spectram which agrees best wdth om' data i n the range around x x 10^ - we find T ,f x t^^ x 1/20 T. Thus the momentum advected away by the wave field is of the same order as the momentum transferred to the low-fi-eciuency waves by noifiinear interactions - or stated differently, the noifiinear momentum fiux to low fi'eciuencies is approximately balanced by that fraction of the momentum input from the atmosphere which remains i n the wave field and is not dissipated. However, for short fetches (xiviO^) the higher JONSWAP a values imply Xy,(Xx so that T , f 1/3 Tijf X 1/3 T, which is approximately sis times greater than T ^^. Hence i n this case the non-dissipated atmospheric input T^^ is unable to support the nonlinear momentum flux to low-frequencies. The momentum gained by the low frecjuencies must therefore be largely compensated by a momentum loss i n the central region of the spectrum, so that - i n contrast to the distribution shown i n Pig. 2.12 - the positive and negative lobes of the net source function S (weighted m t h a factor fe/co to transform from energy to momentum) approximately cancel. A large negative lobe i n S tends to eifiiance the spectral peak relative to the baokground, or increase y while decreasing a. Pig. 2.8 and fom- of the five individual generation cases studied suggest that y does indeed increase for small fetches, wdfile a decreases more rapidly than the power law (2.6.4) for a shape-preserving spectrum. Por fetches x > 10^, the noifiinear momentum transfer to low frecpiencies decreases more rapidly than the net non-dissipated momentum r^^ supplied to the wave field. I t follows that the growdih of the spectrum at low freciueneies is governed by the wind forces (and possiblj' clissipation) rather than wave-wave interactions. I n this case, we should expect the peak to begin to fiatten for x > 10^ Pigme 2.8 and the fom- individual generation cases mentioned above suggest that this may indeed be the case. For very large fetches, the asymptotic Pierson-Mosko-witz spectrum is, of com-se, much fiatter than the mean JONSWAP spectrum. I n summary, we find that for small fetches nearly all of the momentum fiux T across the ah-sea interface enters the wave field i n the central region of the spectrum, i n agreement -ndth P. W. Dobson's [1971] dhect measm-ements of sm-face pressm-e and wave height correlations. About ( 9 0 ^ % of the wave-induced momentum fiux T,îs transferred to shorter waves by
wave-wave interactions, where i t passes to currents through dissipation, while ( ^ g ) " / " * The sensitivity of the noifiinear energy transfer on the spectrum, coupled with the sparsity of reliable data, led K. Hasselmann [1963b] to underestimate the significance of wave-wave scattering for the energy balance of the spectrum. The earlier computations were based on a Neumann spectrum, which has a far less pronomiced peak than observed in the present spectra.
52
Erganzungsheft zin- Deutschen Hydrographischen Zeitschi-ift. Reihe A (8), Nr. 12, 1973
remains i n the -^vave field and is adveeted a-\vay. Approximately (80 + 10) % of the maximal -wave gro-^vth observed on the forward face of the spectrum can be attributed to the nonlinear energy transfer across the spectrum*. The corresponding energy loss on the rear slope of the peak is only partially balanced by the input from the atmosphere and resrdts i n a sharpening of the spectral peak to-wards the equibrium form discussed i n Section 2.8. The energy level of the specti-um adjusts to a value for which the nonlinear interactions are just able to remove (primarily to high wavenumbers) the energy supplied by the wind. With increasing fetch, the nonlinear momentum transfer x^f to high wavenumbers decreases to approximately 15 % of T at a; = 10^. The -\vave gro-svth on the forward face of the spectrum is still dominated by the nonlinear transfer, but the associated energy loss on the rear slope of the peak is no-w largely compensated by the input from the atmosphere. Assuming negligible dissipation i n the main part of the specti-um, approximately 20 % of the momentum transfer to the ocean is needed to maintain the observed energy and momentum balance of the wave field. Alternatively, most of the momentum transfer across the ah-sea interface may be entering into the imnoipal -ivind sea components, as i n the case of small fetches, and is being dissipated i n the same region of the spectrum. Neither possibility can be ruled out on the basis of present observations or theories. The same ambiguity besets the interpretation of the energy and momentum balance i n the remahiing fetch range extending from x = 10^ to the upper l i m i t x = 10*, where the spectral jDarameters border on the values for a fully-developed spectrum - assuming such a spectrum to exist. Our a data, distributed belo-w the mean line (2.4.4), indicate that the nonlinear transfer becomes less important than the atmospheric input i n this range and that the spectral peak should begin to fiatten again. The minimal -ivave-induced momentum fiux without dissipation may fall to as low 5-10 % of the total an--sea momentum fiux - but any fraction up to the theoretical 100 % limit is conceivable i f dissipation i n the central region of the sjiectrum is important. Although we have not been able to observe a systematic transition of om- spectra into a f u l l y developed spectrum of the form proposed by P i e r s o n and M o s k o w i t z , i t is interesting to note that the parameters a = 0.081, = 0.14 of the latter spectrum lie rather close to the mean cmves (2.4.3) and (2.4.4) - as opposed to our a data for 10^ < .-r < 10*. Hence wavewave interactions should still be significant for the fully developed spectrum. Computations of the noifimear momentum transfer r^f to high frequencies for the Pierson-Moskowitz spectrum imply a minimal wave-induced di-ag T " ' " X 0.09 T . (Pig. 2.21b.) 2.10. Variability of the Spectrum Having considered the average properties of the energy balance of the wave field, we return now to the question of the variabfiity of the peak shape parameters. As pointed out in Section 2.4, the considerable scatter of y, and i n Pig. 2.8, and to a lesser extent a i n Pig. 2.7, cannot be explained simply by the statistical uncertainty of the spectral estimates or the neglect of additional external parameters i n K i t a i g o r o d s k i i ' s similarity law. The fact that comparable scatter is observable also vdthin indi-vidual generation eases (Section 2.7) suggests a small-scale mechanism such as the gustiness of the -wind as origin. However, since the peak is a stable feature of the nonlinear energy transfer (Section 2.8), some additional
* Motivated by linear feedback theories of wave generation, the gi-owth rate on the forward face of the spectrum has generally been expressed in previous field experiments i n terms of the exponential growth factor /? = S/E. For a self-simUar spectrum defined by (2.4.1), (2.4.5) we obtain Px6 (]" 10-3 /, where c is the phase velocity and n = 1.03 or 1.33 depending on whether
\ 0J
the scaling laws (2.4.3), (2.4.4) or (2.4.3), (2.4.7) are apphed. Both relations agree within the scatter of the data with R. L . Snyder and C. S. Cox's [1966] and T. P. B a r n e t t and J. 0. Wilkerson's [1967] results in the range 1.2< UJo<Z relevant for this experiment. J. R. Schule et al. [1971] measured somewhat smaller growth rates.
53
instability mechanism must be acting to magnify the effect of weak inhomogeneities of the wind field on the peak shape parameters. An instability of this kind may be identified i n the propagation characteristics of the energy of the peak. Locally, the peak energy propagates i n space at the group velocity corresponding to peak frequency ,. However, at the same time the peak moves in the frequency domain towards lower frequencies. Hence we may envisage the energy of the peak as following a cmwed characteristic i n the x t plane defined by the pah of equations dx dt g inf, m (group velocity at frequency /^)
_ _.;i!a.!Hii:ta;-!l.;t.,'idl
spectrum. I n the ideal case of a homogeneous, stationary wind field, the set of characteristics emanating from the initial (shore) line x = 0 at different initial times are displaced parallel to one another and never intersect. However, small perturbations of the wind field produce local distortions and displacements of the characteristics which normally lead to an intersection of the rays at some distance from the disturbance. A t the caustic points, the peak is enhanced by the superposition of energy from different origins, whereas the energy is reduced i n regions in which the characteristics diverge. ( I t may be noted i n this context that the observed variability of the peak-shaped parameters can not be reduced by considering the combmation y{cr^ + (J,,), which is roughly proportional to the energy i n the peak. Thus the scatter cannot be explained simply as a local redistribution of the energy wdthin the peak, but represents a variability both i n the net energy and the shape of the peak.) Physically, a gust of, say, 2 k m scale w i l l cause a local shfft of the peak towards lower frequencies, causing the peak to propagate faster than peaks generated earlier by the undistm-bed wind at the same fetch. Ultimately, the peak generated i n the gust w i l l then overtake the earlier peaks, (Pig. 2.23). To exjDress this analytically, we need to translate the spectral form of the radiative transfer ecjuation (2.1.1) into an ecjuivalent parametrized form. Let us assume generally that
Fig. 2.23. Focusing of peak propagation characteristics by small scale inliomogeneities of the wind field
54
Erganzungsheft zur Deutschen Hj'chographischen Zeitsclirift. Reihe A (8), Nr. 12, 1973
we can model all spectra F(k; x, t) i n a gdveii problem b j ' a class of fmictions _F'(/f; . . . ft,,) containing n free parameters cij, and that there exists an algorithm for conipntiiig the parameters for any given spectnim F{k) by some best-fitting procedm'e. 2j=cpj{F)
Avliere
(2.10.1) yields a variation of the (2.10.2)
the functional cpj is differeiitiable, i.e. a variation SF{k) parameters Baj= (p'j{8F)
i n which cp'j is a linear functional of SJ^'. Appljdng (2.10.2) to a variation of the model spectrum ~ M= dF 8aj, daj
i t follows that
fdF\ 'P>[-^) = û-
(2-10.3)
Since we assume that F can be approximated by F for all x and t, we may substitute F in the radiative transfer equation (2.1.1) to obtain ^ ( ^ ^ daj \t + . V , ) = ^. J (2.7.1)
Applying the operation (pi and invoking (2.10.3), the parametrized radiative transfer ec[uation takes the form ^ + A . . ' ^ = ^. at o.T^ where Dijk = <p'i (v, ^ ) and SI = (pl {S). (2.10.6) Both the eoeflfioients Dij^ and the source fmiction Sj depend on the parameters a^; the latter also depends on the "wind speed and possibly other parameters. We have neglected the refractive terms i n (2.10.4). I n the ease of the parametrization (fq, . . . a^) = (/,, a, y, a^, (7^) applied i n this experiment, all of the parameters except a depend only on the properties of the spieetrum in a narrow band of frequencies around the spectral peak (we ignore parameters specifying the dhectional distribution, as these were retained constant). Thus (pj and (p! are essentially zero outside this band. I n the functional (p^ in (2.10.5) we may then replace the smoothly varying function V = gl4:uf to first order by the constant value = glinf^, obtaining Dij = fdF\ (p'l I + higher order terms BIJ or, invoking (2.10.3), A v = n, S,j + E,j. (2.10.7) (2.10.8) (2.10.5) (2.10.4)
(The index k has been ch'opped since now x = x.) Equations (2.10.4) and (2.10.8) imply that the parameters characterizing the peak projiagate approximately along characteristics defined by the fetch-dependent group velocity v^, as anticipated.
55
I n the case of a, the appropriate propagation velocities D^j w i l l be smaller than v,,,, as cp^ and cp'^ are governed by the more slowly propagating, shorter waves of the spectrum, a was determined i n the present experiment by least-square fitting a P i e r s o n - M o s k o w i t z spectrum w i t h free ordinate scales a, j to the observed spectrum i n the frequeney range 1.2 to 0.7 Hz (after prewhitening with the factor f ) . W i t h this procedme, a is independent of the shape parameters (which were determined subsequently) and only weakly dependent on so that
(p'Av
=0
for
j = 3, 4, 5
where = (p'^ {v dFjda) is an average group velocity i n the frequency band 1.2 t o 0.7 Hz. Thus for all parameters, the propagation matrix Dfj is approximately diagonal, and eq. (2.10.4) takes the form 9oi at where for foi' i = 1 i = 2, . . . 5 (The index i is enclosed i n parenthesis to indicate f bat i t is excluded from the summation convention.) da, dx dai dx
^^'^ ~ I'^m
Consider now a solution of (2.10.9), a,. = a,--t-a/, which consists of the kno-wn timeindependent solution ; (x) for a given time-independent som-ce S; and a superimposed timedependent fiuotuation a! (x, t) induced by a small fluctuation S'l {x, t). Subtracting the transfer equations for a,, and a, we obtain, -with V(i) = V(i) + (';), S'ij = Wa + (f, da'i dt + da'i dx + da'j _ a. = Rf (a, a', , da ^
X,
, daj ~ ^ (2.10.10)
t), say.
Making use of the fact that the cross-coupled terms 8,.^. da'jjdx are small compared m t h the diagonal propagation terms da'Jdx, equations (2.10.10) may be solved iteratively, the ?ith iteration being defined as the solution of
a^ ^ ' ^ ^
--''^^"-^
a.X,
^ t).
(2.10.11)
+ -Kj [a, _ i ,
Setting a._^ = 0, the first-order solution is obtained by integrating (2.10.11) along the zero'th order characteristics da; = (,)(%) (l=/m)
d9
56
Erganztmgsheft zm- Deutschen Hych-ographischen Zeitscln-ift. Reihe A (8), Nr. 12, 1973
The second iteration then involves integration along the pertiu-bed characteristics do; = vî) ^ = 1 . dS The rays x = x{9,Q, = S + obtained by integrating these equations under the initial condhion x = 0 for t = Avill, i n general, not yield a unique mapping of the enthe x, t plane into the plane of characteristic coordinates 9, since the rays w i l l normally cross at some fimte distance from the initial boundary line x = 0. The caustic linuting the one-to-one mapping region is defined by the condition dx as A = 9a; dt^ dl 39
= V
(1 +
1, i)
, \^v,,,{x(t 0 dtn
9'))d9' = 0.
(2T0.12)
dtg dtg I Beyond the caustic, the solutions g are no longer tmiquely defined. This probably represents a limhation of the iteration procedme rather than a genuine singularity of the complete solution .,. A t the caustic, the pertm-bations ; become large and the iterative approach is no longer valid. This can be seen by continuing the iteration to the t h h d order, for -which the oross-coupled gradient terms on the right-hand side
dcij
daj, = e.
{9, to)
9fn
89 daj, 39 dt.
are found to approach infinity at the caustic A =0. A quantfiative analysis of the behaviour of the solution near the caustic requhes a more detailed investigation of the som-ce function than can be given here. We note here simply that the amplification of the pertatrbations near the caustic is associated formally -with the existence of the cross-coupling gradieifi terms Sjj 8.,/3.rj. I n the case of s f^, these terms vanish since is defined independently of the remaining spectral parameters; the scatter of is indeed fomid to be considerably smaller than for the remaining spectral parameters, a also contains only a -weak cross term e^j- . d f j d x and is observed to have less scatter than the shape parameters, -^vhioh can be sho^wn to be more strongly cross-coupled (particularly to a). Thus intersections of the characteristics yield large variations of the peak energy and bandwidth, but only minor changes i n the peak fi-equency and the high-frequency range of the spectrum - as woifid be expected physically. The distance of the caustics downwind from the generating wind-field inhomogenehies depends on the scale and magnitude of the perturbations. According to eq. (2.10.12), a gust of time scale T causing a local pertm-bation of the peak fi-equency S/ = - i \ f ^ (which then persists along the peak propagation path) vdll result i n the development of a caustic at a downwind distance x defined by 0 = ,^_|9!L.9M.a9' 3 L 9 o or xr T
57
For example, a gust of ten minute dmation producing a 10 % change i n the peak frequency at 0.4 Hz (v^ 2 m/s) woid produce unstable caustic amplification about 12 k m downwind. The values appear consistent with the magnitudes and scales of observed fluctuations (Section 2.7). 2 .11. Conclusions From measm-ements and computations of the net source function, the energy transfer due to nonlinear resonant wave-wave interactions, and the net local momentum transfer from the atmosphere to the ocean, we have been able to identify the following principal featm-es of the energy balance of fetch-limited spectra: 1. The shape of the spectrum is determined primarily hy the nonlinear energy transfer from the central region of the spectrum to both shorter and longer wave components. I n particular, the pronounced peak and steep forward face can be explained as a self-stabilizing featm-e of this process. The rapid wave growth observed for -âves on the forward face of the spectrum can also be largely attributed to the nonlinear transfer to longer waves. 2. For short fetches (.-r = (/.r/C/jo = 0 (10^)) approxhnately 8 0 % + 20% of the total momentum x transferred from the atmosphere to the ocean goes into the principal components of the wave spectrum, i n accordance -with F. W. Dobson's [1971] measurements. About ( + 5\ 5 _ 2 j % of the momentum T transferred fi-om the atmosphere to the waves is retained i n the wave field and is advected away. The rest is transferred by nonlinear interactions to short waves, where i t is converted to current momentum by dissipative processes. I n the main part of the spectrum the dissipation appears to be negligible. 3. For larger fetches {x = 10^ - 10*), the advection of momentum by the wave field is of the order of 5% of the total momentum transfer x, as before, but the momentum transferred to short waves by nonlinear interactions chops to ^20 ^ % of x. I n the absence of dhect measurements of the atmospheric input and the dissipation, i t can be concluded i n this case only that the spectral energy balance must lie between two limiting situations: / + 25\ the wave-induced drag T accounts for ( 25 _ j^g % of T , and all of the dissipation occm-s -via the nonlinear transfer to high wavenumbers, or T is of the order of x, and most of the momentum entering the wave spectrum is dissipated by processes i n v o M n g only a minor redistribution of energy by resonant wave-wave interactions. 4. To first order, wave-taifii and field data from various som-ces scale i n accordance -with S. A. K i t a i g o r o d s k i i ' s [1962] law. Over a five decade range of fetches extending from X = 10-* to 10*, the frequency and energy scales of the wave spectrum vary i n such a manner that the minimal wave-induced drag T " ' " (as inferred from highly scale-sensitive computations of the nonlinear transfer to short waves) lies i n a range between 10 % and 100 % of T supporting the -view that i n general the spectrum adjusts to a level at which the energy input by the wind can be largely removed by the nonlinear energy transfer to shorter waves. This would explain, among other featm-es, the similarity of the spectral shape for wave-tank and field experiments and the observed decrease of Phillips' "constant" a with fetch. The JONSWAP data falls within the last two decades of these plots; the decrease in the ratio T"7T vnth fetch discussed i n points 2. and 3. reflects a small b i f i systematie deviation of the JONSWAP a data from the mean trend over all wave-tank and field data. This may point to limitations i n attempting to scale complex atmospheric boundary-layer processes in terms of the "svind speed alone. A number of questions have remained unanswered. The mechanism of the energy transfer from the -wdnd to the waves was not determined. The transition of the fetch-limited spectrum
58
Erganzungsheft zur Deutschen Hych'ogi-aphischen Zeitscln-ift. Reihe A (8), Nr. 12, 1973
to a f u l l y developed eqiulibrium form (assummg this exists) involves a number of open problems, including the role of dissipation i n the lo-w-frequenoy range of the spectrum. Finally, the mechanism of energy dissipation at high -'avenumbers remains to be clarified. I t is intended to investigate some of these questions i n more detail i n futm-e joint studies. Applied to the problem of numerical wave prediction, the principal conclusion to be cha-\vn from our study is that the nonlinear energy transfer due to wave--(vave scattering requh-es more careful modeUing than hitherto. Ignored enthely i n earlier prediction schemes (e.g. L . Baer [1963], W. J. P i e r s o n , L . J. T i c k and L . B a e r [1966]), the nonlinear transfer terms have been parametrised by rather simple - essentially scaling - relations i n later models (T. P. B a r n e t t [1968], J. A. E-^ving [1971]). I t appears now that a more sophisticated description including a number of critical shape parameters is called for. Appendix Wind field and spectral parameters for generation cases 1-5 as summerised i n Table 2.
59
Profile direction (Offshore, 287)
u = 10m/s [\l
St. km
St. km
10--52
II -- eo
o [5 m/s
0 0 0
\
4-1
27 20
1
20
.1W 3
9.5 6.5
4 O O O O O
15
9.5 6.5
4 2
\ .\.\ A
/
B ^20 ^2
'24
2 J
July 26
1969
July27
'
^tirs 6 y
\
8~\^
.1
10
1
14 \ ^ 16
. \
12 \ ^
Aug 11 1 9 6 8
St. km
10-r52-
St. km.
8 -27 20 15 -26 15
'
/A \
a5
6.5
3?
41
6.5
+5
irs\^ 8 y
l\^
y4
\^
l i
^4
'
Aug 10 9C 6Q 9 vii/i i n1 iQ \ ^ Aug 11969 ' Fig. 2 A.1. History of winds measured along the profile during generation cases 1-5. Dots denote wave measurements, crosses the positions at which the source functions were computed
\ \
fTATTu
60
61
10"
10"
Fig. 2 A.3. Shape parameters, and ^ ;r for generation case 1. Meaningful x and \ derivatives cannot be defined on account of the scatter
Fig. 2 A.4. Same as Fig. 2 A.2 for generation case 2
Hasselmann e t a l . J O N S W A P
64
65
.25r
10
10"
10
10'
IO-'
66
Fig. 2 A.8.
Same as Fig. 2 A.2 for generation case 4
Hasselmann e t a l . J O N S W A P
Fig. 2 A.9. Same as Fig. 2 A.3 tor generation case 4
68
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr.12,1973
Fig. 2 A.10.
Same as Fig. 2 A.2 for generation case 5
69
3. Swell Attenuation 3.1. Conservation of Action and Energy Mux A l i t t l e less than one-thu-d (654) of the wave spectra recorded duidng JONSWAP contained well-defined low-frequency swell peaks. I n most cases the swell energy decreased from the outer to the imiermost stations of the profile by faetors varying from 0.2 to 0.7; the strongest decay occm-red along the inshore, shallower section of the profile (cf. Fig. 3.1). The swell beams were sufficiently narrow to be characterized simply by their mean frecpiency /,
0.12
h 0630 jui
5)
0.10
(FROM S W E L L CASE
5 0.08
0.06
0O4 h
0.02
0.1
0,2
0.3
0.4
0.5
0.6
0.7
Fig. 3.1. Examples of sweh peaks observed dm-ing swell case 5. Numbers refer to wave stations. The swell energj^ decreases towards shore, with the exception of Station 1, where the energy is eiihanced by refraction (the energy flux I at Station 1 is generally smaller than at Station 2, cf. Fig. 3.2)
72
Erganzimgsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12, 1973
frequency width A/, mean dhection 0 ( = , cf. Section 1.5), directional spread A0 {= 0^) and total energy - or, more conveniently, the total shoreward energy flux / = v^, where = V cos 0 is the group velocity of the swell i n the direction parallel to the profile. I n the ideal case of no energy loss, negligible refraction by cmrents, parallel bottom contours orthogonal to the profile and constant swell frequency, the fiux / remains constant along the propagation path of the swell. I n the general case i n which no restrictions are invoked other than the standard assumption of slowly varying mean fields, the rate of change of I or S can be deduced from the action balance equation + (.-i-.iV) + [kN) dt dki = Slco' (3.1.1)
where N{k; x, t) =F{k; x, t)lco' is the action spectrum, co' = [gk tanh kH)''''' is the (intrinsic) frequency of the wave components i n a reference system moving with the local cm-rent lJ{x, t), H{x, t) is the water depth, X; = = dQjdkj , dQ oo--,and (0 = Q{k, cc, t) = {gk tanh kHY''^ + k-U denotes the dispersion relation i n a fixed reference system. Ecj[uation (3.1.1) represents the generalization of the energy balance equation (2.1.1) to the case of refi-action by arbitrary space- and time-dependent ciu-rents and water depths. I t expresses i n spectral form the conservation of action for indi-vidual wave packets (cf. G. B. W h i t h a m [1967], P. P. B r e t h e r t o n and C. J. R. G a r r e t t [1968]) propagating along the rays i n x, k phase space defined by eqs. (3.1.2). Noting that ^ form (rf;,.) + ~(ki) DN = 0 (eq. (3.1.2)), eq. (3.1.1) may be-svi-itten i n the equivalent , , , (3.1.3) (3.1.2)
dN dN ^ dN 8 = + Xi + ki = T)t dt dxi dki co'
which states that i n the absence of an energy som-ce 8 the action density remains constant along the path of a wave group in x, k space ( L i o u v i l l e ' s theorem, ef. R. D o r r e s t e i n [1960], G. E. B a c k u s [1962]). I n applying (3.1.3) to the case of a narrow swell peak, we may integrate over the small region of wavenumbers representing the swell to yield a propagation equation for the total swell energy <, ,+ dt \co'J dxi \ = co'J co' (3.1.4)
where $ = j 8dk and the swell frequency co' was taken as constant i n the integration. I n the particular case i n which all fields are homogeneous -\idtli respect to y, the em-rents vanish, and the frequency remains constant, eq. (3.1.4) reduces (after multiplying by co'jv^ and noting that dvjdt = 0 for c = const) to the simple form ^ ^ 1 9 ^ + A z = ^ Dx dt dx (3.1.5)
where D\Dx represents the total derivative with respect to the normal-to-shore coordinate x 1 V taken along the propagation path of the swell - i.e. vdth df = dx, dy = dx (the latter variable is hrelevant i n the present context, however). ^-^
73
The routine analysis of om- s-\vell data -\vas based on the idealized energy flux equation (3.1.6). Ho-wever, integrations of the exact equation (3.1.4) weve also carried out for a number of special cases, using a general ray refraction routine -^vbich included spatial and temporal variations of both cm-rents and "water depths. The errors incurred i n (3.1.5) through neglect of cm-rents, tidal variations of -water depth and slo^w changes of the s-well fL-ec[uency "were found to be negligible. Significant deviations from the ^/-independent case arose, ho-wever, through the local focusing caused by lateral variations of the bottom topography. The modifications were found to be highly sensitive to small variations of the swell frequency and dhection and A v e r e -vh-tually impossible to model deterministically (Section 3.4). 3.2. Attenuation Due to Bottom Fi'iction I n the generation study (Part 2), the som'ce function i n the energy balance equation was determined emphically by differentiating the observed spectral fields. We adopt here the inverse procedure and integrate the energy flux ecpation (3.1.5), assuming a pai-ticular form for the source function. The theoretical model can then be tested by comparing the predicted Avith the observed spectra. This approach avoids the inaccuracies of experimental differentiation, but is feasible only i f the som-ce functions and spectra are sufficiently simple to permit a straightforward numerical integration. I n the present case, the simplicity stems from the narrow peak approximation and our preconceptions regarding the structm-e of the source function (w-hich w i l l be found subsequently to need revision). I t is generally assumed that the attenuation of swell propagating i n shallow^ water is caused by bottom friction. Wave tank measm-ements (J. A . P u t n a m and J. W. J o h n s o n [1949], P . P . Savage [1953], K . K a j i u r a [1964], Y . I w a g a k i et al. [1965]) and - to a lesser extent - field data (C. L . B r e t s o h n e i d e r and R. 0. R e i d [1954], H . W a l d e n and H . J. R u b a c h [1967], K . H a s s e l m a n n and J. I . C o l l i n s [1968]) support the usual assumption that the tangential stress T ,, acting on the ocean bottom can be modelled approximately by a cpiadratie friction law x^ = cÛ\U\ (3.2.1)
where U denotes the cm-rent immediately outside the t h i n tm-bulent boundary layer at the bottom and c,, is a drag coefficient, which i n the ocean is generally of order 10"^. K . H a s s e l m a n n and J. I . C o l l i n s [1968] have shown that a friction law of the form (3.2.1) implies a cpasi-linear som-ce function (cf. eq. (2.1.1)) S = Sî{k)=
- V i j k k j F { k )
(3.2.2)
i n which the coefficient Vjj represents a kinematic -viscosity tensor given by
The expressions i n cornered parentheses denote ensemble averages and can be computed from the known Gaussian distribution of the cmrents. Substituting eqs. (3.2.2) and (3.2.3) i n (3.1.1) and integrating over the swell peak, we obtain the net source function = - ^ ^ ( < | r . | > + co^cosh^ kH\ (f^J) \\U\/ J (3.2.4)
where C7// denotes the component of U parallel to the direction of wave propagation. The above relations apply for an arbitrary bottom cm-rent U = U^ + consisting of a superposition of the wave orbital velocity t/^^ and a mean cm-rent U". I n most of om- swell cases, the wave orbital velocities varied -svithin the range 0.1-5 cm/s, whereas the tidal currents were typically of order 20^10 cm/s. Hence to a good approximation the expectation
74
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8"), Nr. 12, 1973
values on the right hand side of eq. (3.2.4) may be replaced by the appropriate expression for the mean cm-rent. Assummg constant V along a ray, the flux equation (3.1.5) can then be readily integrated to give lnJ=c,^F^ (3.2.5) where V = \V'\ + {Ufifl\U'\ (3.2.6) and = ] (o^ cosli^ kH Dx. (3.2.7)
According to (3.2.5) the energy flux / decays exponentially -ivith respect to the normalized distance coordinate ^ at a rate proportional to the current factor V. 3.3. Energy Flux Analysis o Swell Data Ten distinct swell cases were investigated (Table 3). For each swell case, the decay of the energy flux / -with respect to the normalized distance ^ -was computed for a sequence of rays at fom--hourly intervals. Figs. 3.2a-j (to avoid congestion, only every second or t h u d decay curve is sho-ivn i n some of the figm-es). The points on the curves correspond to station positions, the flux values at the times corresponding to the transition of the rays thi-ough the stations being obtained by interpolating between the four-hom-ly measm-ements at each station. The origin of ^ is chosen at Station 1, and the coordinate increases "with distance x offshore. Due to the cosh-^ kH weighting factor i n the definition (3.2.7), most of the change i n ^ occurs i n the shallow water close to shore, resulting i n a more or less linear increase of , wdth respect to station number. Table 3 Swell Cases (No. of independent spectra: 678-24= 654) of No. of No. Spec- Rays tra (4 lu-. seciu.) 12 Sep. 24 Sep. 24
5
Swell Case
Start Time
End Time
8/
r-103
1 2 3 4 5 6 7 8
9
1968 121' Sep. 8 G O H Sep. 24 091 Sep. 24

0411
231 231
10-3 Hz lO-sHz deg
s-i
m 2 s-3
2 3 4 5 13 7 5 17 8 9
89 86-91 88-102 78-98 102-116 77-120 119-137 68-98 87-101 80-92
14-17 10-17 10-16 10-17 10-20 6-15 9-16 8-16 8-15 6-14
107 120 120 120 130 141 138 99 120 120
2-10-2
10-1
160 45 80 49 34 1.9 89 27 43 30 Mean: 38
24 37 36 142 80 53 179 56 59
10-1-2-10-1
jal, 4
1969 021 Jul.
10-1-2-10-1
2011 Jul. 14
121
221 Jul. 16
141
2-10-2-4-10-2 5-10-3-2-10-2
3-10-1
10-2-5-10-1 10-3-3-10-3 5-10-1-10-3
Jul. 22 Jul. 26 Aug. 2

9
Jul. 23 Jul. 30 Aug. 10
0411
221 Jul. 26
091
101 Jul. 27
1511
021 Aug. 4
191
10
081 Aug.
Total: Total: 678 73
75
10
15
^(sec^m'^)
10
410''
10
15
C (sec^m^)
76
to
-3 h
Fig. 3.2 e-g.
Same as Fig. 3.2 a-d
77
Fig. 3.2 ti-j. Same as Fig. 3.2 a-d
70
X > 5 POINTS <4P0INTS

0.3
APPROX. ERROR BANDS
0.2
^to 0.1
95%
CONF
CONF
^)5%
-0.2
0.2 0.4 0.6 0.8
V(m/s)
957o
-0.2
-4
-2
HRS, RELATIVE
TO HIGH TIDE
HELGOLAND
Fig. 3.3 a,b. Decay parameterT versus current parameter V and phase of tide. Data are divided into two classes x, depending on number of energy flux measurements / available to define the mean slope F (eq. (3.3.1)). Stepped lines denote 95 % confidence limits for moan value? of T in finite abcissa segments
79
X > 5 POINTS < 4 POINTS APPROX. E R R O R BANDS 957o
10
12
14
95% CONF
1
X K 95% CONF.
-0.1
=12
-4
O UfQCOS<p(m/s)
Fig. 3.4 a,b. Decay parameter T versus wind speed and component of wind velocity parallel to swell propagation direction. Data are divided into two classes X , depending on number of energy flux measurements / available to define the mean slopeT (eq. (3,3.1) )
80
X2 5 POINTS s 4 POINTS ^ APPROX. ERROR I , BANDS I .95%

0.2
0.3
com
h
0.1 Hi*
co\w.
.5
0
-0.1 -0.2
1
I
10'
JO-
10-
10^
I(m^sec')
0.3
0.;^
0.1
j
95% COWF. 95% COiMR
ii
ul
=0.1 = -0.2 0D6
008
0.10
//Z
0.12
Fig. 3.5 a,b. Decay parameterF versus swell frequency and energy flux. Data are divided into two classes x, depending on number of energy flux measurements / available to define the mean slope F {eq. (3.3.1))
Hasselmann et al., JONSWAP To eaeli decay ciu've a straight line was fitted by least squares. According to eq f3 2 5) the slopes of the lines i ^ />
r dln/ ^ = ^
(3.3.1)
should be equal to c^^gV, so that a marked time variation was to be expected through the dependence on the thne varymg cm-rent factor V. However, a cursory inspection of Tigs. 3.2a-j does not reveal an obvious tidal periodicity. This is borne out by the plot of the set of all computed T values against V, Tig. 3 3a which shows no dependence of V on T, as predicted by (3.2.5). The values of V were iirferred for Station 8 from tidal em-rent charts, which were found generally to give a good approximation to the measm-ed cm-rents. Errors inem-red by assuming a constant cm-rent along the ray were also relatively small. The lines represent 95% confidence bands for the average values of r defined for separate 5 om/s segments of the F axis. The statistical basis of these esthnates is described i n the Appendbc. (Note that the r.m.s. error for the average within a segment is considerably smaller than for the individual estimate of T, as is apparent by the narrowness of the band.) A general test of tidal influences on the decay rates, independent of the particular mechanism assumed, is provided by the plot of T against the phase of the tide. Fig. 3 3b The tune axis represents the thne lag of the arrival of the swell at Station 1 relative to the nearest high tide i n Helgoland (a convenient reference point i n the vicinity of the profile cf. Fig. 2.2 - for which published tide tables exist). The representation is'meaningful only i f the tides are predominantly semidim-nal, but this is the case. Again, no tidal modidation is apimrent. Figm-es 3.4a, b indicate that the swell decay rate is also essentially independent of the local -nund speed and the component of the whid velocity parallel to the propagation dhection of the swell. A dependence on these variables may have been anticipated either dhectly thi-ough ah-sea interactions or indhectly through the (WKBJ-type) coupling bet-O'een the swell and the shorter -svind-wave components (0. M. P h i l l i p s [1963], K . H a s s e l m a n n [1971]). Ho-svever, neither process appears to be important i n the present situation. Finally, the dependence of the decay rate on the swell parameters themselves was investigated. No systematie variation was found m t h the swell propagation dhection (of Table 3) or the swell frequency (Fig. 3.5a). However, there is some indication of an increase of the decay rate with the swell energy or, equivalently, the energy flux / (Fig. 3.5b). Desphe the apparent limhations of the bottom friction model, i t is perhaps of interest that the enhanced decay rates for / ^ 1 m^ s-i are consistent with this theory. For 1 = 1 m^ s-i the r.m.s. orbhal velocity near the bottom is of the order 10-20 cm/s, and since this is comparable m t h the tidal currents U, the approxhnation <g leading to equations (3.2.5), (3.2.6) is no longer valid. The inclusion of the orbhal velochies increases the value of the expression ( < |7| > + < UÎ\U\>) on the right hand side of (3.2.4) and thereby the attenuation factor T. For I x 10 r is enhanced by a factor of approxhnately 2, as observed. ' 3.4. lateral Variations of Bottom Topography Om- analysis of the attenuation parameter T i n the previous section was based on the /-independent energy flux equation (3.1.5). Lateral variations of the depth modify these results i n two ways. Fhstly, the normalized distance ^ as deflned by eq. (3.2.7), is no longer a unique function of x, since dhferent depth profiles correspond to different ray paths. Computations of ^ using the true bottom topography and ray paths for a number of swell cases indicated that the eiTors induced i n T i n this manner were generally less than 20%. Secondly, the flux equation (3.1.5) hseff is no longer strictly valid. The missing lateral energy flux terms ( i n addition to time dependent variations) are taken account of formally i n the
82
Erganzimgsheft ziu- Deutschen Hydrograpliisohen Zeitschriffc. Reihe A (8"), N r . 12, 1973
general action balance equation (3.1.4). However, the straightforward application of (3.1.4) is restricted to regions i n which the rays do not intersect. I n practice i t was found that i n nearly all the cases computed the profile was intersected by at least one caustic and that the ray theory therefore needed to be generalized to allow for multiple sheets i n the x, y plane. Although this is feasible i n theory, i t was foimd, i n practice, that the cav;stics were produced by small scale irregularities of the bottom topography and their locations varied unstably with small changes i n frecûency and propagation du-eotion of the swell. Apart from the position of the caustics, the ray integrations revealed generally that most of the variability of the energy flux was induced by small scale topographic features and was w t u a l l y impossible to predict deterministicalty. The predominance of small scale features is illustrated i n Fig. 3.6, which shows examples of depth sections along the profile and along a ray thi'ough Station 8 at an incident angle of 34 to the profile.
STATIONS
DISTANCE OFFSHORE (km)
F i g . 3.6. B o t t o m depths along wave array (echo somider) and a ray t h r o u g h Station 8 (computed f r o m depth charts). The mean depth profile assumed i n the parallel-taottom-contour calculation is also shown
A comparison of ty^Dical sets of rays computed for the idealized case of parallel bottom contours and the true bottom topographj^' are shown i n Fig. 3.7. I t is apparent that small local pertm-bations can produce significant changes i n the ray pattern at large distances from the distm-bances. According to elementar}' ray theory, a depth pertm-bation Si? of dimensions orthogonal to and L// parallel to the direction of wave propagation will produce focusing at a distance ^^lXlsinh(2fc//) 2 LuUH
behmd the distm-bance. For a wavelength of 150 m (period 10 s) and typical values (see cf. Fig. 3.6) Si = 2 m , i j ^ = i / / = 4 k m , 7? = 15 m, we obtain ? 30 km. Closer to shore, the focusing distances are generally smaller due to the reduced value of the factor smh {2kH). Thus i t appears that the observed variabilitj' of the energj^ fiuxes i n Figs. 3.2a-j can be explained cj[ualitatively b j ' the convergence and divergence of rays due to small scale bottom u'regularities. (A similar mechanism was suggested as the origm of the variability of the spectral parameters observed i n the generation study. Part 2.) A n attempt was made to detect a possible systematic mean modification of the energy flux superimposed on these fluctuations by computing the refi-action patterns for a smoothed bottom topography. Fig. 3.8 shows typical energy fluxes computed from the ray patterns for a bottom topography i n which all scales smaller than 10 k m were filtered out. The somxe
83
Fig. 3.7. Rays computed for the true bottom topography and the parallel bottom-contour approximation (0,08 Hz)
84
Fig. 3.9 a-e.
Mean swell frequencies versus time for swell cases 3 -10. Tidal modulation of this frequency bandwidth for swell case Different curves correspond to different stations
85
86
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr.12,1973
Rg. 3.9 h,i. Same as F\g. 3.9 a-e
87
function was talin as zero, so that variations o f ! are due enthely to lateral inhomogeneities. The values still fluctuate considerable from station to station and change randomly with propagation dhection. Corresponding curves for neighbouring frequencies also differed appreciably. Although i n this case no caustics occm-red, i t appears that the refraction pattern is still dominated by relatively short topographic features of scale slightly larger than the cut-off applied by the fllter. A systematic dependence of the energy flux with respect to station number, dhection of propagation or frequency could not be detected. The observed energy fluxes (Figs. 3.2a-j) generally exhibit smaller variability than Fig. 3.8 - which applies to a monochromatic wave held - would lead one to expect. Presumably, the flux values for a real swell fleld should be obtained by integrating the monochromatic fluxes over a finite band of frecjuencies and propagation dhections, wlrieh averages out some of the variability. Similar conditions of random refraction may apply generally i n continental-margin areas, i n which the scale of the topographic features are usually small compared with the propagation distances over which the swell can interact significantly with the bottom. 3.5. Dispersion Characteristics I n t h e h original study of swell off the coast of Cornwall, N . F. B a r b e r and F . U r s e l l [1948] observed a characteristic linear increase of the swell frequencies with time which they explained i n terms of a singular-event model. I f the space-thne region i n which the swell is generated is small compared -with the swell propagation distances and times, the source can be approximated by a <5-function. Accordingly, only those swell components w i l l be observed by an observer at distance D from the source for which D = group velocity X travel time, or f= l i ^ (3.5.1)
where t is the time of origin of the swell. W. H . M u n k and F. E. Snodgrass [1957], W. H . M u n k , G. R. M i l l e r , F. E . Snodgrass et al. [1963], F. E. Snodgrass et al. [1966], D. E. C a r t w r i g h t [1971], and others have applied (3.5.1) to determine the distances and time of occurrence of swell som-ces, i n general findhig excellent agreement with expected source regions as determined from weather maps. The investigations were concerned with relatively long-period swell, normally i n the 12-20 second range, originating i n storms at great distances - i n some cases exceeding half the ohcumference of the earth. I n the present study, dhect access to the open ocean was restricted to the small northern opening of the North Sea. The longest available wave path, from the east coast of Greenland, was only 3000 km, but even this was limited to the 10" mndow between Iceland and Scotland. Consequently, most of the swells observed were generated rather near to the wave stations, and the dimensions of the som-ce regions were normally not significantly smaller than the swell propagation distances. Moreover, all swell periods lay i n the range between 7 and 15 s (frequently near 12 s), indicating that the swells were generated by rather moderate winds (10-23 m/s), which normally persist locally for longer periods than intense storms. Both factors combined to make the application of the usual (5-event model c[uestionable in many of om- swell cases. Accordingly, the frec[uenoy histories shown i n Figs. 3.9a-i often reflect the space-time evolution of the generation region as strongly as dispersion effects, and the relation (3.5.1) cannot be applied i n a straightforward manner to determine the origin of the swell. However, by considering the propagation characteristics of the waves i n conjunction with the weather maps, the various swell sources shown i n Fig. 3.10 and theh "case histories" could be reconstructed "with some confidence. Swell oases 4, 6 and 7 could be approximated by the 5-event model. The thne origins and distances inferred from (3.5.1) were identified on the weather maps with rather localized, short-lived regions of high winds directed towards Sylt. Case 1 probably also belongs to this
88
ErganzLuigsheft zm- Deutschen Hydrographischen Zeitschrift. Eeihe A (8), Nr. 12, 1973
30
20*
10
10
Fig. 3.10. S-well sources as inferred from weather maps and wave propagation times category, but the swell was observed too briefly to determine the dispersion rate. The same restriction applies to swell case 2. Swell eases 8, 9 and 10 show positive dispersion, but here the som-ce distances inferred from (3.5.1) were either impossible ( > 3000 km, cases 8 and 10) or not consistent with the weather maps (case 9)*. However, i n these cases the observed dispersion could be explained natm-ally by the spatial and time dependence of less distant generating regions identified on the weather charts. For example, dmdng the swell period 8 - the longest observed - the suspected generation region gradually wandered from south-east of Iceland into the northern part of the North Sea, weakemng as i t moved. Thus the high-frec|uency components generated later but nearer to the wave stations dominated over the swell components generated earlier at greater distances. The net effect was an anomalously slow frequency increase with time. Similar explanations apply to swell cases 9 and 10. Converse^, the negative freciuency trend of swell case 3 was attributed to a gradual increase of the -ndnd strength i n a fahly stationary generation region west of Norway. Swell case 5, finally, "U'as associated -ndth a stationary wind system north of Scotland. The narrowness of the swell peaks sho-wn for this case i n Fig. 3.1 were apparently not caused by dispersion, but simply refiect the peakedness of the spectrmn i n the generating region. Consistent with this interpretation, the ratios peak-Avidth to peak-frequency are seen to be approximately the same for both the swell peaks and the higher-frequency, locally generated regions of the spectra. Swells which persisted for a day or longer exhibited a pronounced semi-diurnal modulation of the freqixency. N . F. B a r b e r and F. U r s e l l [1948] first noted this effect and attributed i t * Beflections off Greenland or Iceland in cases 8 and 10 were also inconsistent with the weather maps.
89
to the Doppler shift by tidal eurrents. Clianges i n the swell frequenej' are governed by the equation (ef. eq. (3.1.2)) Bco _dQ 'Dt~ 8t~ 8U dt
where the left hand side represents differentiation along the path of a wave group. I n most cases, the swell travel time was of the same order as a tidal cycle, so that Aoj x k U, where U is the mean tidal current. Taking ^ ~ and Aco 27t x 0.003 sî we obtain
X 0.5 m/s, which is a reasonable value for the North Sea. Interestingly, the frequencj' bandwidth of the swell peak often exhibits a stronger tidal modulation than the peak fL'ec[uency itseh, cf. T i g . 3.9g. 3.6, Conclusions We have not been able to identify clearly the mechanism of swell attenuation. Interpreting the mean decay rate F = 0.04 m^ s"^ i n terms of a bottom friction model, assummg the usual quadratic stress relation, and taking a mean value 25 cm/s for the current parameter V, we obtain a drag coefficient = 0.015. This is consistent wdth previous estimates. However, the theory predicts a linear dependence of the decay rate on V, which was not observed. Nor could a tidal modulation of the decay rate be detected. Although the quadratic friction law can undoubtedly only be regarded as a crude approximation of the dissipative processes in a non-stationary turbulent boundary layer, i t appears improbable that the swell deeaj? rate shordd be virtually independent of the rather strong tidal cm-rents i f bottom friction is indeed the dominant decay mechanism. The search for correlations vnth. other parameters was ec[ually unsuccessful. Neither the wind speed nor the component of the wind velocitj? parallel to the dhection of -wave propagation affected the decay rate. No dependence was found on the swell frecjueney or propagation dhection. However, there was a marginal indication of an increase of the decay rate with the swell energy. Theoretical considerations rule out wave-wave interactions (K. H a s s e l m a n n [1963b]) or interactions with turbulence outside the bottom boundary layer (K. H a s s e l m a n n [1968]) as significant mechanisms of swell attenuation. Recalling that F. E. Snodgrass et al. [1966] observed no measm-able attenuation of Pacific swell over distances of many thousand kilometers, and that the strongest decay rate i n the present study was observed close to shore, some f o r m of bottom interaction still appears the most lUicly candidate. Although improbable, i t is conceivable that the interactions with the mean current i n the bottom boundary layer are such that the rate of wave energy dissipation happens to t m n out independent of the mean cm-rent - despite the fact that the mean current greatly exceeds the wave orbital velocities. A more attractive alternative is baokscattering due to Bragg interactions with bottom irregidarities of scale comparable -with the wavelength of the swell. Recent computations by R. L o n g [1972] indicate that bottom "waves" i n this wavelength range of only 10-20 cm r.m.s. amplitude would be adequate to explain the observed attenuation. However, detailed spectral measm-ements of the bottom topography and the separation of the incident and backseattered components of the swell field, both of which were not carried out i n the present study, are needed to test this hypothesis. W i t h respect to swell kinematics, classical dispersive swell systems generated by distant storms appear to be the exception rather than the rule i n the North Sea. The frecjfueney histories of most swells were characterised by signatures impressed by the particiar spacetime sti-ucture of the generating wave fields, Variations of the swell energy by factors of two or more are common and can be understood through the ray focusing caused by lateral inhomogeneities of the bottom topography. The refracting features are generally of relatively small scale; the cumulative
90
Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Beihe A (8), Nr. 12, 1973
effect of many independent refractions produces a randomly scattered swell beam, i n analogy with the scintillation of celestial objects seen through the atmosphere and other small-angle scattering problems (cf. V. I . T a t a r s k i [1961], L . A. C h e r n o v [I960]). Appendix Statistical Analysis of the Attenuation Parameter F The least-sc[uare-error regression line through 7i measm-ements Zj = In I j at points ^j, 7 = 1, . . . w is given by z = F{^-0+-z (A.1) where F = {z--z)(^l)k^ - If (A.2)
and the over-bars denote mean values with respect to the ensemble of data points j = 1, . . . n. We make the standard assumption that errors i n the estimate of F arise through errors i n the variable ( j . = ( Z j - Zj) {^j - ^j) vnth fixed ^j, and that the variance al = (C]} of (j w i t h respect to a hypothetical ensemble of repeated experiments is independent of j. I n this case, F is approximately Gaussian for large n and the variance of F approaches
4
where
= " If-
(A-3)
al = {Cm
Ecp (A.2) was used to determine F for all swell rays k. I n plotting these values F^ against various other parameters, such as the tidal current F, the tidal phase, swell frequency, etc., the Fj^ data i n separate segments of the parameter axes were combined into a single estimate
m
T = E c , n
k= l
(A.4)
where k=i, .. . m refer to the in values i n a given segment. The weights c ^ , were chosen to minimize the variance al
= <(f -
iFyfy
= S 4 <(r, - <r^>
(A.S)
subject to the side condition

m
E c , = l.
(A.6)
Substituting (A.3) i n (A.5) (and replacing n now by ij) the minimal condition becomes
I; ^ = 0
k=l
%
subject to (A.6), which has the solution

n
c,, =
The variance of F then becomes

2
al=
T~
m k=l
(A.7)
91
Eq. (A.7) was applied to eompute the 95% confidence limits shown i n Figs. 3.3-3.5. On account of the repeated sum definitions (A.2) and (A.3), F is fairly closely normal, and the 95 % confidence bandwidth is apjDroximately equal to 4:aj. The quantity i n (A.7) was estimated using (A.3). Multiplying the equation by replacing < ( r > by the approximate vahie T, and then summing over the m points of the interval, we obtain
m
mal
~S
k= l
(T, - F f .
This equation was then averaged over all intervals, yielding finally
2 "O ~ 2 _
|%(A-r)4
j
intervals h = l
î:,
intervals
Notation page spectral parameter 42 drag coefficient 12 co-spectrum of time-series i and j 20 bottom friction coefficient 12 equivalent friction coefficients 41 cb-ag coefficient corresponding to T^^, 50 minimal cb-ag coefficient corresponding to T " 50 coefficients i n parametrical radiative transfer ec|uation 54 variance of smface displacement (total "energy" of the -svave field) 38 nondimensional variance ("energy") of surface displacement 38 one-dimensional spectrum 20 Pierson-Mosko-îdtz spectrum 32 energy spectrum with respect to propagation direction 0 and frequency? / 19 two-dimensional wave spectrum A v i t l i respect to wave nmnber k 40 model spectrum containing 7i free parameters a^, . . . . a 54 frequency 11 non-cUmensional frequency parameters 28 out-off-frequency 41 frequency of spectral peak 15 non-dimensional frec[uency of spectral peak 32 water depth 15 total shoreward energy flux 72 k = (kj^, k^) horizontal wavenumber vector 20 modulus of k 20 wave number component kj, k^ averaged over directional distribution 20 length dimensions orthogonal and parallel to the direction of wave propagation 82 Nj = N(kj), N = N(k) particle-number or action density at wavenumber kj,k 40 exponents i n power laws for a and S versus non-dimensional fetch 38 exponent charactering directional distribution 32 quadi-ature speetrimi of time series i and j 20 coherence 21 one-, two-dimensional som-ce functions 27 som-ce function due to bottom friction 73 source function representing dissipative processes 49 source fimction representing energy input from the atmosphere 49 som-ce function due to non linear, resonant wave-wave interactions 49 source function of parametrical radiative transfer equation 55 normalized angular distribution 20 wind velocity at 10 meters height above the sea surface 28 mean current at bottom 73 orbital velocity at bottom 73 current 72 component of U parallel to the direction of wave propagation 73 mean tidal cm-rent 89
a, CJQ C f j (or C) C|, Cj ,j, c^f c ^ ^ c"^" îj' îjk ( # E (/) ^ P M (/) F(f,6) p (It) F{k, . ... a,|) /', /, /
f^, H / k k kj, k^ L^, Lji Bp N HJ-, n^, 11 ~ p Qij (or Q) R 8,8' (S'bf
JS'JJ
;Si 8^1 8i s{f, 6), ov s(9) U" C J w U{x, t) Uli TJ
Hasselmann et al., JONSWAP friction velocity cm-rent parameter group velocity component averaged over directional distribution = ( 1 , 2) group velocity group velocity corresponding to frequency of spectral peak average group velocity i n the frequency band used to define a (cUmensional) fetch nondimensional fetch parameters
93 n 74 27 27 37 55 H 11
V Vj V v X x', X, X a ^ y fiij' fia/m I 6 0ni 0s 0m> d's S A, A, Vij Q^^ i7a, (Tb T Tad Tb Ti,f Tif Tn,f T T" <Pj, <Pj Q CO co'
Phillip's equilibrium constant 32 exponential gro-wth factor 52 ratio of the maximal spectral energy to the maximum of the corresponding Pierson-Mosko-^vitz spectrum 32 couphng coefficients i n propagation equations for spectral parameters 54 normalized s-ivell propagation distance 74 direction of "wave propagation 20 mean dhection of -wave propagation 21 beam -\vidth defined i n terms of piteh-roh directional spectrum 21 variables analogous to 0 ^ and 0^ for a pah- of -svave sensors 21 time variable of characteristic coordinates 55 -^vavelength, -tvavelength at spectral peak 15 kinematic viscosity tensor 73 density of air 27 density of -neater 40 left, right sided m d t h of spectral peak 32 total momentum transfer across the air-sea interface 27 momentum advected aAvay by the vf&ve field 41 tangential bottom stress 73 net momentum transfer corresponding to the third (high frequency) lobe of the nonlinear source function 41 net momentum transfer corresponding to the first (lo-w freciuency) lobe of the noifiinear source function 51 net momentum transfer corresponding to the middle (negative) lobe of the nonlinear source function Si 51 net momentum fiux from the atmosphere to the wave field 12 lower bound of the momentum transfer from the atmosphere to the wave field 49 functional, hnear functional, derivative of (pj 54 dispersion relation 72 circular frequency i n fixed coordinate system 72 frequency i n a reference system moving -ndth the current U {x, t) 72
References
Backus, G. B., 1962: The effect of the earth's rotation on the propagation of ocean waves over long distances. Deep-Sea Res. 9, 185. Baer, L . , 1963: An experiment m numerical forecasting of deep water ocean waves. Srmnyvale: Lockheed Missile Space Co., LMSC-801296. Barber, N . F. and F. U r s e l l , 1948: The generation and propagation of ocean waves and swell. 1. Philos. Trans, r. Soc. (A) 240, 527. B a r n e t t , T. P., 1968: On the generation, dissipation, and prediction of wind waves. J. geophys. Res. 73, 513. B a r n e t t , T. P., 1970: Wind waves and swell in the North Sea. Eos, Wash. 51, 544. B a r n e t t , T. P. and A. J. Sutherland, 1968: A note on an over-shoot effect in windgenerated waves. J. geophys. Res. 73, 6879. B a r n e t t , T. P. and J. C. W i l k e r s o n , 1967: On the generation of ocean wind waves as inferred from ahborne radar measm'ements of fetch-limited spectra. J. mar. Res. 25, 292. Becker, G., K. P. K o l t e r m a n n , G. P r a h m und W. Z enk [in prep.]: Die hydrographische Situation in der nrclliohen Deutschen Bucht im Juh 1969 (JONSWAP). Dt. hych'ogr. Z. B r e t h e r t o n , F. P. and C. J. R. G a r r e t t , 1968 : Wave trains in inhomogeneous moving media. Proc. R. Soo. (A) 302, 539. ^Zg 2 Bretsohneider, C. L . and R. O. Reid, 1954: Changes in wave height due to bottom friction, percolation, and refraction. Techn. Mem. US Ai'my Eros. Bd. No 45, 36 pp. Brocks, K . and L . K r g e r m e y e r , 1970: The hydrodynamic rouglmess of the sea surface. Ber. Inst. Radiomet., Hamburg. No. 14. (Also i n : Studies in physical oceanography. Ed. by A. L . Gordon. New York. 1, 75.) C a r t w r i g h t , D. E., 1971: Tides and waves in the vicinity of Saint Helena. Philos. Trans, r. Soc. (A) 270, 603. C a r t w r i g h t , D . E . and N. D. S m i t h , 1964: Buoy techniques for obtaining directional wave spectra. Buoy Teclmology. Washington: Marine Teclmology Society. Chernov, L . A., 1960: Wave propagation in a random medium. New York: McGraw-Hill. D a r b y s h i r e , J., 1963: The one dimensional wave spectrum in the Atlantic Ocean and in coastal waters. I n : Ocean wave spectra. Prentice Hall. Page 27. Davis, R. E., 1969: On the high Reynolds numljer flow over a wavy boimdary. J. Fluid Mech. 86, 337. Davis, R. E., 1970: On the turbulent flow over a wavy boundary. J. Fluid Mech. 42, 721. D e L e o n i b u s , P. S., 1971: Momentmn flux and wave spectra observations from an ocean tower. J. geophj's. Res. 7G, 6506. Dobson, F. W., 1971: Measurements of atmospheric pressure on wind-generated sea waves. J. Fluid Mech. 48, 91.
D o r r e s t e i n , R., 1960: Simplified method of determining refraction coefficients for sea waves. J. geophys. Res. 65, 637. E l l i o t t , J., 1972: Microscale pressm'e fluctuations near waves being generated by the wind. J. Fluid Mech. 54, 427. Enke, K., 1973: Turbulenzmessungen ber See bei begrenztem Fetch. Diplomarbeit, Inst. Geophys., Univ. Hambiug. E w i n g , J. A., 1971: A numerical wave prediction method for the North Atlantic Ocean. Dt. hydrogr. Z. 24, 241. Hasse, L., 1968: On the determination of the vertical transports of momentum and heat in the atmospheric boundary layer at sea. Hambm'ger Geophys. Einzelsclu'. Nr. 11. (Also: Techn. Rep. Dep. Ooeanogr., Ore. St. Univ. No. 188, Ref. No. 70-20. 1970.) Hasselmann, D. E., 1971: Wave generation by resonant casemode interactions in a turbulent wind. Trans. Amer. geophys. Un. [in prep.]. Hasselmann, K., 1962: On the non-linear energy transfer in a gravity-wave spectrmn. 1: General theory. J. Fluid Mech. 12, 481. Hasselmann, K., 1963a: On the non-linear energy transfer in a gi'avity-wave spectriun. 2: Conservation theorems, wave-particle correspondence, irreversibility. J. Fluid Mech. 15, 273. Hasselmann, K., 1963b: On the non-linear energy transfer in a gravity-wave spectrum. 3: Computation of the energy flux and swell-sea interaction for a Neumami spectrum. J. Fluid Mech. 15, 385. Hasselmann, K., 1966: Feynman diagrams and interaction rules of wave-wave scattering processes. Rev. Geophys. 4, 1. Hasselmann, K., 1967: Nonlinear interactions treated by the methods of theoretical phj^sics (with application to the generation of waves by wind). Proc. R. Soc. (A) 299, 77. Hasselmann, K., 1968: Weak-interaction theory of ocean waves. I n : Basic developments in fluid dynamics (Ed. M . Holt). New York. 2, 117. Hasselmann, K., 1971: On the mass and momentum transfer between short gravity waves and larger-scale motions. J. Fluid Mech. 50, 189. Hasselmann, K., 1972: The energy balance of wind waves and the remote sensing problem. Conf. Sea Surface TopograjDhy, Miami, October, 1971 [in press]. Hasselmann, K . and J . I . Collins, 1968: Spectral dissipation of finite-depth gravity waves due to turbulent bottom friction. J. mar. Res. 26, 1. I w a g a k i , Y., Y . T s u c h i y a and M . Sakai, 1965: Basic studies on wave damping due to bottom friction. 2: On the measm'ement of bottom shearing stress. Bull. Disast. Prev. Res. Inst. Kyoto Univ. 14, 45.
Hasselmann et al., .JONSWAP J e f f r e y s , J., 1924: On the formation of waves by wind. Proc. R. Soc. (A) 107, 189. Jonsson, I . G., 196.5: Friction factor diagrams for oscillatorjr bomidar3r layers. Basic Pes. Progr. Pep. coast. Engng Lab. Teclin. Univ. Denmark. No. 10, 10. K a j i u r a , K., 1964: On the bottom friction in an oscillatory current. Bull. Earthq. Res. _ Inst. 42, 147. K i t a i g o r o d s k i i , S. A., 1962: Applications of the theory of similarity to the analysis of wind-generated wave motion as a stochastic process. Bull. Acad. Sci. USSR Geophys. Ser. No. 1, 73. K o r v i n - K r o u k o v s k y , B. V., 1967: Pmther reflections on properties of sea waves developing along a fetch. Dt. hydrogr. Z. 20, 7. L i u , P.O., 1971: Normalized and equilibrium spectra of wind waves in Lake Michigan.
J. phj 'S. Oceanogr. 1, 249.
95
Long, R., 1971: On generation of ocean waves by a turbulent wind. Ph. D . Thesis, Univ. Miami. Long, R., 1972: Scattering of surface waves by bottom irregularities, (subm. to J. geophj^s. Res.) L o n g u e t - H i g g i n s , M. S., D. E. C a r t w r i g h t and N . D. S m i t h , 1963: Observations of the directional spectrmn of sea waves using the motions of a floating buoy. I n : Ocean wave spectra. Prentice Hall. Page 111. L o n g u e t - H i g g i n s , M. S. and Pv. W. Stewart, 1964: Radiation stress in water waves, a physical discussion with applications. DeepSea Res. 11, 529. Miles, J. W., 1957: On the generation of surface waves by shear flows. J. Fluid Mech. 3, 185. Blitsuyasu, H . , 1968a: On the growth of the spectrum of wind-generated waves. 1. Bep. Res. Inst. Appl. Mech., Kyushu Univ. 16, 459. Blitsuyasu, H . , 1968b: A note on the nonlinear energy transfer in the spectrum of wmd-generated waves. Rep. Res. Inst. Apjal. Mech., Kyushu Univ. 16, 251. M i t s u y a s u , H . , 1969: On the growth of the spectrum of wind-generated waves. 2. Rep. Res. Inst. Appl. Mech., Kyushu Univ. 17, 235. M i t s u y a s u , H . , R. N a k a y a m a and T. K o m o r i , 1971: Observations of the wind and waves in Hakata Bay. Rep. Res. Inst. Appl. Mech., Kyushu Univ. 19, 37. Munk, W. H., G. R. M i l l e r , F. B. Snodgras.s and N . F. Barber, 1963: Dh-ectional recording of swell from distant storms. Philos. Trans, r. Soc. (A) 255, 505. Munk, W. H . and F. E. Snodgrass, 1957: Measurements of southern swell at Guadalupe Island. Deep-Sea Res. 4, 272. P h i l l i p s , O. M., 1958: The equilibrium range in the spectrmu of wind-generated ocean waves. J. Fluid Mech. 4, 426. P h i l l i p s , O. M., 1963: On the attenuation of long gravity waves by short breaking waves. J. Fluid Mech. 16, 321. P h i l l i p s , O. M., 1966: The dynamics of the upper ocean. Cambridge: University Press.
Pierson, W. J., Jr. and L . M o s k o w i t z , 1964: A proposed spectral form for fully developed wind seas based on the similarity theory of S. A. Kitaigorod.skii. J. geophys. Res. 69, 5181. Pierson, W. J., Jr. and R. A. Stacy, 1973: On the elevetation, slope and ourvatme spectra of a wind roughened sea [in press]. Pierson, W. J., Jr., L . J. T i c k and L . Baer, 1966: Computer based procedure for preparing global wave forecasts and wind field anal3rsis capable of using wave data obtained by a space craft. 6th Symp. Naval H5i-drogr., Washington, Offico of Naval Research. Pond, S., G. T. Phelps, J. E. Paquin, G. McBean and R. W. Stewart, 1971: Measurements of the turbulent fluxes of momentum, moisture, and sensible heat over the ocean. J. atmos. Sci. 28, 901. P u t n a m , J. A. and J. W. Johnson, 1949: The dissipation of wave energ3? hy bottom friction. Trans. Amer. geophys. Un. 30, 67. Ross, D. B., V. J. Cardone and J. W. Conaway, 1970: Laser and micro-wave observations of sea-surface conditions for fetchlimited 17- to 25-m/s winds. IEEE Trans., Geo.sci. Electronics GE-8, 326. Savage, R. P., 1953: Laboratory study of wave energy by bottom friction and percolation. Techn. Mem. U.S. Ai-my Eros. Bd., No. 31, 25 pp. Schule, J. J., L . S. Simpson and P. S. D e L e o n i b u s , 1971: A study of fetchlimited wave spectra with an airborne laser. J. geophys. Res. 76, 4160. Sell, W. andK. Hasselmann, 1972: Computations of nonlinear energy transfer for JONSWAP and emph-ical wind wave spectra. [Rep.] Inst. Geophys., Univ. Hamburg. S m i t h , S. D., 1967: Thrust-anemometer measurement of wind-velocity spectra and of Rejmolds stress over a coastal inlet. J. mar. Res. 25, 239. S m i t h , S. D., 1970: Thrust-anemometer measurements of wind turbulence, Reynolds stress and di-ag coefficient over sea. J. geophys. Res. 75, 6758. Snodgrass, F. E., G. W. Groves, K . F. Hasselmann, G. R. M i l l e r , W. H . M u n k and W. H . Powers, 1966: Propagation of ocean swell across the Pacific. Philos. Trans, r. Soc. (A) 259, 431. Snyder, R. L . and C. S. Cox, 1966: A fleld study of the wind generation of ocean waves. J. mar. Res. 24, 141. Sutherland, A. J., 1968: Growth of spectral components in a wind-generated wave train. J. Fluid Mech. 33, 545. T a t a r s k i , V. I . , 1961: Wave propagation in a turbulent medium. New York: McGraw-Hill. Walden, H . und H . J. Rubach, 1967: Gleichzeitige Messungen des Seegangs mit nichtstabilisierten Beschleuniguugsschreibern an Orten mit miterschiedlicher Wassertiefe in der Deutschen Bucht. Dt. hydrogr. Z. 20, 157. W h i t h a m , G. B., 1967: Variational methods and applications to water waves. Proc. R. Soc. (A) 299, 6.
Bisher erschienen die folgenden Erganzungshef te zur Deutschen Hydrographischen Zeitschrift

Reihe A (8")
Nr. 1: Walter Hansen,
Gezeiten tnid Gezeitenstrme der halbtagigen Hauptmondtide WI^ in der Nordsee (1952, Preis 4,50 DM) Nr. 2: Georg Koopmann, Entstehung und Verbreitmig von Divergenzen in der oberflachennahen Wasserbewegung cler antarktischen Gewasser (1953, Preis 5,50 DM) Nr. 3: Heinz Gabler, Nautische Teohnik (Formeln, Diagramme, Tabellen) (1955, Preis 2 3 , - DM) Nr. 4: Hans Theis, ber erchiiagnetische Pidsationen (1957, Preis 10, DM)
Nr. 5: Dt. Hydrogr. Inst., Dt. P'orschungsg'enieinschat (Hrsg.), Forschtmgssohiff MeNr. 6: Klaus Wyrtki, Nr. 7: Aloys Heupel, Nr. 8: Fritz Model,
Nr. 9: Friedrioh Scbott, Nr. 10: EiUiard Cordes, Nr. 11: Kayihan Arip,
teor" (1964, Preis 3 , - DM) The Thermal Structm'e of the Eastern Pacific Ocean (1964, Preis 5 , - DM) Neue Wege bei der Herstelhmg deutscher Seekarten (1965, Preis 4,80 DM) Geophysikalische Bibliographie von Nord- und Ostsee (in 3 Banden). Teil I : Chronologische Titelaufzahlung. Band 1: 1749-1932. Band 2; 1933-1961. Teil I I : Register. Berlin: Borntraeger i . Komm. (1966, Preis 4 8 , - DM) Der Oberflachensalzgehalt in der Nordsee (1986, Preis 14,40 DM) Die LiteraturersclrlieBung in der Meereskunde (1970, Preis 8,60 DM) Der Krustenaufbau und die Tiefenstruktur des ReykjanesRckens, sdwestlich von Island, nach reflexionsseismischen Messmrgen (1972, Preis 1 6 , - DM)
Reihe B (4)
Hydrographische Untersuchungen in der Ostsee 1925 bis 1938 mit dem Reichsforsclimigsdampfer ,,Poseidon" (1956, Preis 1 2 , - DM) Magnetische Reichsvermessmig 1935. 0. Teil I I (Karten) Nr. 2: B . Bock, F . Burmeister, F . Errulat, (1956, Preis 1 2 , - DM) Die Expeditionen von F.F.S. Anton Dohrn" und V.F.S. Nr. 3: G. Bhnecke, ,,Gauss" im Internationalen Geophysikalischen Jahr 1957/ A. Bckmann (Hrsg.), 1958 (1959, Preis 1 8 , - DM) Nr. 4: G. Dietrich (Hrsg.), Temperattu-, Salzgehalts- mid Sauerstoffverteihmg auf den Sclmitten von F.F.S. Anton Dolu-n" mid V.F.S. Gauss" im Internationalen Geophysikalischen Jahr 1957/1958 (1960, Preis 9, DM) Trbungs- unci Temperatm'verteilung auf den Stationen und Nr. 5: J . Joseph, Schnitten von V.F.S. ,,Gauss" sowie Bathythermogramme von F.F.S. ,,Anton Dolirn" und V.F.S. ,,Gauss" im Internationalen Geophysikalischen Jain- 1957/58 (1961, Preis 9,60 DM) Echolotprofile der Forschungsfalu'ten von F.F.S. ,,Anton Nr. 6: J . Ulrich, Dohrn" und V.F.S. ,,Gauss" im Internationalen Geophysikalischen Jalrr 1957/1958 (1962, Preis 1 8 , - DM) Monatskarten der Temperatur der Nordsee, dargestellt fur Nr. 7: G. Tomczak, verschiedene Tiefenhorizonte (1962, Preis 9, DM) E . Goedecke, Die thermische Schiohtung tier Nordsee auf Grund des mittNr. 8: G. Tomczak, leren Jahresganges der Temperatur in mid 1"-Feldern E . Goedecke, (1964, Preis 19,60 DM) Monatskarten des Salzgehaltes der Nordsee, dargestellt fr Nr. 9: E . Goedecke, J . Smed und G. Tomczak, verschiedene Tiefenhorizonte (1967, Preis 1 8 , - DM) Seekarten der sdlichen Nord- tmd Ostsee. Ihre Entwicklung Nr. 10: A. W. Lang, von den Anfangen bis zmn Ende des 18. Jalu'hunderts. Berlin, Stuttgart: Borntraeger i . Komm. (1968, Preis 56, DM) Monatskarten der Temperatur der Ostsee, dargestellt fr verNr. 11: W. Lenz, schiedene Tiefenlaorizonte (1971, Preis 28, DM) Monatskarten des Salzgehaltes cler Ostsee, dargestellt fr verNr. 12: K . - H . Bock, schiedene Tiefeidiorizonte (1971, Preis 28, DM) Monatskarten der Dichte des Wassers in der Ostsee, dargestellt Nr. 13: K . - H . Bock, fr verschiedene Tiefenhorizonte (1971, Preis 28, DM)
Nr. 1: Bruno Schulzf,

Hasselmann Etal DHZ1973

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Hasselmann Etal DHZ1973

Diunggah oleh

Hak Cipta:

Format Tersedia

Measurements of Wind-Wave Growth and Swell Decay during the Joint North Sea Wave Project (JONSWAP)

Erganzungsheft zur Deutschen Hydrographischen Zeitschrift Reihe A (8), N r . 12

1973 DEUTSCHES HYDROGRAPHISCHES INSTITUT HAMBURG

Erganzungsheft zur Deutschen Hydrographischen Zeitschrift

DEUTSCHES HYDROGRAPHISCHES INSTITUT HAMBURG

DEUTSCHES HYDROGRAPHISCHESINSTITUT HAMBURG

Ansohriften der Vertasser

Hasselmann et al., JONSWAP

Hasselmann et al., JONSWAP

Erganzungsheft zur Deutscfien Hydrograptiischen Zeitschrift. Reihe A (8), Nr. 12,1973

Hasselmann et al., JONSWAP

infiuenced appreciably by the bottom; the wavelength

ST. 5 JULY 2 2 1 9 6 9 2 0 0 0 X WAVE R I D E R + WAVE F L O A T

ST. 5 JULY 2 2 1 9 6 9 1 6 0 0 X WAVE R I D E R + WAVE F L O A T

ST. 4 JULY 2 2 1 9 6 9 1 8 3 0 X PITCH ROLL WIRE

ST. 4 JULY 2 2 1 9 6 9 1 6 3 0 X PITCH ROLL WIRE

Fig. 1.4 a-d. Intercomparison measurements of frequency wave spectra

POWER SPECTRUM (m^/Hz)

St. 5 Sept 6 19681230

Hasselmann et al., JONSWAP

2D' 0 WAVE STAFF e PRESSURE TRANSDUCER D=7m D'=4D = 28m

Hasselmann et al., JONS^VAP

i - R ' -, {kr cos 0 ; ) ^

1 .1 1 where the coherence

where / j is a normalismg parameter, making

24 Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973

Hasselmann et al., JONSWAP

D UPPER ANEMOMETERl UNSTABLE LOWER ANEMOMETERl STRATIFICATION

L I U (1971) W E I L E R e BURLING (1967)

2.H LARGE KRAUS (1968) FETCH

Fig. 2.2. Reynolds stress sprectrum [k = 2ii U L/,,

Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973

Hasselmann et al., JONSWAP

10^ m/s 3-13 6-15 3-11 3-13 3-15 4-7 3-11

neutral stable neutral unstable stable unstable

profile correlation correlation correlation

E (/) = ag' {2n)-*

exp ( - ; { ^ Y ) y ' " ^

Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973

PEAK SHAPE PARAMETERS

0.0 10^ 10-

Hasselmann et al., JONSWAP

Hasselmann et al., JONSWAP

Hasselmann et al., JONSWAP

Erganzungsheft zur Deutschen Hydrographischen Zeitschrift. Reihe A (8), Nr. 12,1973

Table 2 Generation Cases

15 Sep. 1968 26 Jul. 1969 1 Aug. 1969

m/s 7.0 8.8 9.0 10.7 9.2

9.1 15.6 15.5 20.4 20.5

Hz 0.263 3.06 0.371 0.421 0.349 3.67 3.42 3.86

83 104 92 77 100 90 70 123

0.36 0.046 -0.24 0.045

0.08 0.15 0.09 0.22 0.18

0.13 0.20 0.14 0.29 0.25 0.18

0.333 15.4 18.3 18.0 18.5

0.325 0 + 0.05 0.421 0.411 0.428 0.16 -0.27 0.12 0

0.636 0.386 0.685 0.381 0.608 0.390

0.18 0.052 0 + 0.02 0.072 0.16 0.072 0

1 Aug. 1969 10 Aug. 1969

0.416 3.24 3.3 4.2 7.0

0.050(3) 0.13 0.12

> 20- 0.14 50

(3) From equation (2.4.7)"

Hasselmann et al., JONSWAP